Advanced GNSS Processing Techniques (Working Group 1)
André Garcia Sá2019, Advanced GNSS Tropospheric Products for Monitoring Severe Weather Events and Climate
visibility
…
description
169 pages
link
1 file
Over the last decade, near real-time analysis of GPS data has become a well-established atmospheric observing tool, primarily coordinated by the EIG EUMETNET GPS Water Vapour Programme (E-GVAP) in Europe. In the near future, four operational GNSS will be available for commercial and scientific applications with atmospheric science benefiting from new signals from up to 60 satellites observed at any one place and time, however, many challenges remain regarding their optimal combined utilization. Besides raw data streaming, recent availability of precise real-time orbit and clock corrections enable wide utilization of autonomous Precise Point Positioning (PPP), which is particularly efficient for highrate, real-time and multi-GNSS analyses. New GNSS constellation signals, products and processing methods suggest the development of advanced GNSS tropospheric products, in support of weather
Related papers
Precise near real-time GNSS analyses at Geodetic Observatory Pecnýprecise orbit determination and water vapour monitoring
Acta Geodyn. Geomater, 2010
This paper provides a summary of the Geodetic Observatory Pecný achievements within the Centre for the Earth Dynamics Research (CEDR, 2005(CEDR, -2009 project in the field of precise near real-time GNSS analyses. The GOP data centre supporting our own near real-time activities as well as those of various other institutes has been enhanced by including GLONASS data, real-time GNSS data and some other supporting products. The ultra-rapid GNSS orbits are routinely determined and predicted at GOP by analysing a global network of 60 stations. Significant improvements, which resulted in the fitted and predicted satellite position rms of 4 and 10 cm, respectively, were achieved within the CEDR project. The GOP orbit product is updated every 6 hours and it routinely contributes to the International GNSS Service (IGS). Based on these predicted precise orbits, the GOP near real-time regional GNSS network solution is routinely provided for monitoring water vapour in the atmosphere. Resulting zenith troposphere delays achieved a standard deviation of 3-5 cm compared to precise EUREF post-processing results or a standard deviation of 1-2 mm when converted to precipitable water vapour and compared to a nearby radiosonde. The troposphere delays estimated in GOP are operationally used in the Numerical Weather Prediction.
Real-time national GPS networks: Opportunities for atmospheric sensing
Earth, Planets and Space, 2000
Real-time national Global Positioning System (GPS) networks are being established in a number of countries for atmospheric sensing. UCAR, in collaboration with participating universities, is developing one of these networks in the United States. The network, named "SuomiNet" to honor meteorological satellite pioneer Verner Suomi, is funded by the U.S. National Science Foundation. SuomiNet will exploit the recently-shown ability of groundbased GPS receivers to make thousands of accurate upper and lower atmospheric measurements per day. Phase delays induced in GPS signals by the ionosphere and neutral atmosphere can be measured with high precision simultaneously along up to a dozen GPS ray paths in the field of view. These delays can be converted into total electron content (TEC), and integrated water vapor (if surface pressure data or estimates are available), along each GPS ray path. The resulting continuous, accurate, all-weather, real-time upper and lower atmospheric data create a variety of opportunities for atmospheric research. In this letter we describe SuomiNet, its applications, and the opportunity to coordinate national real-time GPS networks to create a global network with larger scientific and operational potential.
Real-time national GPS networks for atmospheric sensing
Journal of Atmospheric and Solar-Terrestrial Physics, 2001
Real-time national global positioning system (GPS) networks are being established in a number of countries for atmospheric sensing. The authors, in collaboration with participating universities, are developing one of these networks in the United States. The proposed network, named "SuomiNet" to honor meteorological satellite pioneer Verner Suomi, is funded by the US National Science Foundation to exploit the recently shown ability of ground-based GPS receivers to make thousands of accurate upper and lower atmospheric measurements per day. Phase delays induced in GPS signals by the ionosphere and neutral atmosphere can be measured with high precision simultaneously along a dozen or so GPS ray paths in the ÿeld of view. These delays can be converted into integrated water-vapor (if surface pressure data or estimates are available) and total electron content (TEC), along each GPS ray path. The resulting continuous, accurate, all-weather, real-time GPS moisture data will help advance university research in mesoscale modeling and data assimilation, severe weather, precipitation, cloud dynamics, regional climate and hydrology. Similarly, continuous, accurate, all-weather, real-time TEC data have applications in modeling and prediction of severe terrestrial and space weather, detection and forecasting of low-latitude ionospheric scintillation activity and geomagnetic storm e ects at ionospheric mid-latitudes, and detection of ionospheric e ects induced by a variety of geophysical events. SuomiNet data also have potential applications in coastal meteorology, providing ground truth for satellite radiometry, correction of synthetic aperture radar data for crustal deformation and topography studies, and detection of scintillation associated with atmospheric turbulence in the lower troposphere. In this paper we describe SuomiNet, its applications, and the larger opportunity to coordinate national real-time GPS networks to maximize their scientiÿc and operational impact.
Use of GNSS Tropospheric Products for High-Resolution, Rapid-Update NWP and Severe Weather Forecasting (Working Group 2)
2019
For more than a decade, GNSS-meteorology has been increasingly used operationally in Europe particularly for data assimilation in Numerical Weather Prediction (NWP) models, mainly thanks to the EIG EUMETNET GNSS Water Vapour Program (E-GVAP, 2005-today). As such, GNSS has become a well-established, mature observing technique for data assimilation applications. Over this period however, scientists and specialists in GNSS-meteorology noted the clear potential for enhancements and novelties in the domain. The work carried out by the COST Action ES1206 Working Group 2 members addressed these potential enhancements and novelties from the meteorological point of view, in collaboration with WG1. This included the establishment of discussion channels with forecasters in order to determine which GNSS products would be best suited for their day-to-day operational requirements. Particular areas of interest include engaging more operational forecasters (e.g. use of meteorological case studies),...
Module 3 Modèle pour les Données non Normalisées Passage à la technologie objet de Oracle Modèle Objet (Oracle) Modèle objet-relationnel (OR = relationnel + objet
Le modèle géré par Oracle est dit Objet-Relationnel parce qu'il peut faire cohabiter les modèles relationnel et objet . Il est hybride !! Seule la partie exclusivement objet sera utilisée. Avantages:
A Historical Archaeology of the Ottoman Empire
Contributions To Global Historical Archaeology, 2002
Preface the process, we hoped to break down some of the barriers which isolated scholars working in various regions throughout the former imperial provinces. The feedback and responses provided results which were far more rewarding than what we initially set out to accomplish. Charles E. Orser, Jr. encouraged us to create this volume, and Eliot Werner and Herman Makler at Kluwer Academic/Plenum Publishers helped us carry it through to completion. We appreciate their patience and support throughout the various phases of this project, since a number of seasons of field work in Turkey and Israel led to delays. We have been pleased to see interest in an archaeology of the Ottoman Empire develop over the past decade. Yet, despite a growing interest in archaeology of the Ottoman Empire, we still are concerned that this field has yet to find its place in the ranks of more established archaeological research. We hope that this volume demonstrates the potential of archaeological investigations for understanding the recent past of the Middle East, and introduces an archaeology of the Ottoman Empire to a wider scholarly audience. As we discuss in the introduction, there are many avenues for an archaeology of the Ottoman Empire. We envision the volume as an invitation to a dialogue within the field of archaeology on the Ottoman period material record and encourage discussions of the theoretical implications of the case studies, If this volume encourages archaeologists to place the Ottoman period within their preview and consider some of its material remains, we would be satisfied with the endeavor. UZI BARAM LYNDA CARROLL Contents
LA VIDA CONVENTUAL EN LA VILLA DE SAN CLEMENTE
Artículo de divulgación que debió acompañar a un vídeo para la promoción de los conventos de la villa de San Clemente (Cuenca)
Chapter 3
Advanced GNSS Processing Techniques
(Working Group 1)
J. Douša, G. Dick, Y. Altiner, F. Alshawaf, J. Bosy, H. Brenot,
E. Brockmann, R. Brožková, Z. Deng, W. Ding, K. Eben, M. Eliaš,
R. Fernandes, A. Ganas, A. Geiger, G. Guerova, T. Hadaś, C. Hill,
P. Hordyniec, F. Hurter, J. Jones, M. Kačmařík, K. Kaźmierski, J. Kaplon,
P. Krč, D. Landskron, X. Li, C. Lu, J. P. Martins, G. Möller, L. Morel,
G. Ófeigsson, R. Pacione, C. Pikridas, E. Pottiaux, J. Resler, W. Rohm,
A. Sá, J. Sammer, T. Simeonov, W. Söhne, A. Stoycheva, A. Stürze,
Sz. Rozsa, F. N. Teferle, S. Thorsteinsson, P. Václavovic, H. Valentim,
B. Van Schaeybroeck, P. Viterbo, K. Wilgan, L. Yang, L. Zhao, N. Zinas,
and F. Zus
In the following sections material is republished with kind permission: 3.2.1, 3.3.2–3.3.7, 3.4.3,
3.4.6, 3.4.7. 3.4.8, 3.5.3–3.5.6, 3.5.8, 3.5.9, 3.6.2 and 3.6.6.
J. Douša (*)
Geodetic Observatory Pecný, RIGTC, Ondřejov, Czech Republic
e-mail: [email protected]
G. Dick
GFZ German Research Centre for Geosciences, Helmholtz Centre Potsdam, Potsdam, Germany
e-mail: [email protected]
Y. Altiner · W. Söhne · A. Stürze
BKG, Federal Agency for Cartography and Geodesy, Frankfurt, Germany
e-mail: [email protected]; [email protected]; andrea.stuerze@bkg.
bund.de
F. Alshawaf · Z. Deng · X. Li · C. Lu · F. Zus
GFZ German Research Centre for Geosciences, Potsdam, Germany
e-mail: [email protected]; [email protected]; [email protected];
[email protected]; zusfl[email protected]
J. Bosy · T. Hadaś · P. Hordyniec · K. Kaźmierski · J. Kaplon · K. Wilgan
Wrocław University of Environmental and Life Sciences, Wrocław, Poland
e-mail: [email protected]; [email protected]; [email protected].
pl; [email protected]; [email protected]; [email protected]
H. Brenot
Royal Belgian Institute for Space Aeronomy, Uccle, Belgium
e-mail: [email protected]
E. Brockmann
Swiss Federal Office of Topography swisstopo, Wabern, Switzerland
e-mail: [email protected]
© Springer Nature Switzerland AG 2020
J. Jones et al. (eds.), Advanced GNSS Tropospheric Products for Monitoring Severe
Weather Events and Climate, https://doi.org/10.1007/978-3-030-13901-8_3
33
34
J. Douša et al.
R. Brožková
Czech Hydrometeorological Institute, Prague, Czech Republic
e-mail: [email protected]
W. Ding · F. N. Teferle
University of Luxembourg, Luxembourg, Luxembourg
e-mail: [email protected]; [email protected]
K. Eben · P. Krč · J. Resler
Czech Institute of Computer Science, Academy of Sciences, Praha, Czech Republic
e-mail: [email protected]; [email protected]; [email protected]
M. Eliaš · P. Václavovic · L. Zhao
Geodetic Observatory Pecný, Research Institute of Geodesy, Topography and Cartography,
Zdiby, Czech Republic
e-mail: [email protected]; [email protected]; [email protected]
R. Fernandes · H. Valentim
University of Beira Interior, Covilhã, Portugal
e-mail: [email protected]; [email protected]
A. Ganas
National Observatory of Athens, Athens, Greece
e-mail: [email protected]
A. Geiger · F. Hurter
ETH Zurich, Zürich, Switzerland
e-mail: [email protected]; [email protected]
G. Guerova
Physics Faculty, Department of Meteorology and Geophysics, Sofia University “St. Kliment
Ohridski”, Sofia, Bulgaria
e-mail: [email protected]fia.bg
C. Hill · L. Yang
University of Nottingham, Nottingham, UK
e-mail: [email protected]; [email protected]
J. Jones
Met Office, Exeter, UK
e-mail: jonathan.jones@metoffice.gov.uk
M. Kačmařík
Institute of Geoinformatics, VŠB Technical University of Ostrava, Ostrava, Czech Republic
e-mail: [email protected]
D. Landskron · G. Möller · J. Sammer
Department of Geodesy and Geoinformation, TU Wien, Wien, Austria
e-mail: [email protected]; [email protected]; julia.
[email protected]
J. P. Martins · P. Viterbo
Instituto Português do Mar e da Atmosfera, Lisbon, Portugal
e-mail: [email protected]; [email protected]
L. Morel
École Supérieure des Géomètres et Topographes, Le Mans, France
e-mail: [email protected]
3 Advanced GNSS Processing Techniques (Working Group 1)
35
G. Ófeigsson · S. Thorsteinsson
The Icelandic Meteorological Institute, Reykjavík, Iceland
e-mail: [email protected]; [email protected]
R. Pacione
e-GEOS/Centro di Geodesia Spaziale-Agenzia Spaziale Italiana, Matera, MT, Italy
e-mail: [email protected]
C. Pikridas
Aristotle University of Thessaloniki, Thessaloniki, Greece
e-mail: [email protected]
E. Pottiaux
Royal Observatory of Belgium, Brussels, Belgium
e-mail: [email protected]
W. Rohm
Institute of Geodesy and Geoinformatics, Wrocław University of Environmental and Life
Sciences, Wroclaw, Poland
e-mail: [email protected]
A. Sá
Polytechnic Institute of Guarda, Guarda, Portugal
e-mail: [email protected]
T. Simeonov
Sofia University “St. Kliment Ohridski”, Sofia, Bulgaria
e-mail: [email protected]
A. Stoycheva
National Institute of Meteorology and Hydrology, Sofia, Bulgaria
e-mail: [email protected]
Sz. Rozsa
Budapest University of Technology and Economics, Budapest, Hungary
e-mail: [email protected]
B. Van Schaeybroeck
Royal Meteorological Institute of Belgium, Uccle, Belgium
e-mail: [email protected]
N. Zinas
Tekmon Geomatics, Ioánnina, Greece
e-mail: [email protected]
Abstract Over the last decade, near real-time analysis of GPS data has become a
well-established atmospheric observing tool, primarily coordinated by the EIG
EUMETNET GPS Water Vapour Programme (E-GVAP) in Europe. In the near
future, four operational GNSS will be available for commercial and scientific
applications with atmospheric science benefiting from new signals from up to
60 satellites observed at any one place and time, however, many challenges remain
regarding their optimal combined utilization. Besides raw data streaming, recent
availability of precise real-time orbit and clock corrections enable wide utilization of
autonomous Precise Point Positioning (PPP), which is particularly efficient for highrate, real-time and multi-GNSS analyses.
New GNSS constellation signals, products and processing methods suggest the
development of advanced GNSS tropospheric products, in support of weather
36
J. Douša et al.
numerical prediction and nowcasting will be substantially improved. Such examples
are: ultra-fast and high-resolution tropospheric products available in real-time or on a
sub-hourly basis, parameters monitoring tropospheric anisotropy above the station
(such as horizontal gradients and tropospheric slant path delays), and indicators of
severe weather such as extreme convection. Development of advanced GNSS
tropospheric products within COST Action ES1206 benefited from two dedicated
campaigns prepared for a collaborative effort: (1) the benchmark campaign and
(2) the real-time demonstration campaign. The former served for estimating and
assessing horizontal tropospheric gradients and tropospheric slant delays, estimated
from GNSS, Water Vapour Radiometers and Numerical Weather Model (NWM)
ray-tracing. The second campaign developed new software and strategies for realtime, multi-GNSS, high-rate tropospheric solutions including the assessment of
pre-operational solutions.
The impact of selected processing strategies and precise models were assessed
during a long-term GNSS reprocessing campaign aimed at providing homogeneous
tropospheric products for climate research. Using information from modern NWM
forecasting systems, a variety of tropospheric correction models for real-time kinematic GNSS positioning were developed and assessed. Finally, a transfer of knowledge such as support for establishing new GNSS Analysis Centres and inclusion of
new networks into E-GVAP were completed.
3.1
Introduction
J. Douša
Geodetic Observatory Pecný, RIGTC, Ondřejov, Czech Republic
e-mail: [email protected]
G. Dick
GFZ German Research Centre for Geosciences, Helmholtz Centre Potsdam,
Potsdam, Germany
e-mail: [email protected]
The GNSS4SWEC Working Group 1 (WG1) focused on development and utilization of advanced GNSS processing techniques for the purpose of both estimation and
exploitation of tropospheric parameters within geodetic, meteorological and climate
applications. The main goals of the WG1 are summarized within four defined
domains:
1. Coordinating the development of advanced GNSS tropospheric products in
support of weather forecasting, namely such as real-time parameter estimation,
troposphere asymmetry monitoring and modelling, developing severe weather
indicators, assessing the impact of hydrometeors, advantage of multi-GNSS data
processing.
2. Exploiting NWM data in GNSS precise positioning and real-time kinematic
applications, namely supported by mapping functions or mapping factors, a priori
separation of hydrostatic contributions, tropospheric horizontal gradients, tropospheric parameter scaling and conversions factors.
3. Reprocessing of GNSS data and assessing precise models for the purpose of a longterm consistent tropospheric product provision in support of a climatology research.
3 Advanced GNSS Processing Techniques (Working Group 1)
37
4. Supporting the transfer of knowledge, tools and data exchange for extending
existing products and establishment of new ACs or inclusions of new networks.
WG1 coordination was split into ten different sub-tasks, with many overlaps. For
this reason, the structure of this chapter does not correspond to the WG1 sub-tasks,
but reflects mainly goals specified above. An important role in new development was
achieved by a collection of common data sets and the design of specific campaigns,
which are described in Sect. 3.2. Development of advanced tropospheric products
specified in the first goal are introduced in Sects. 3.3 and 3.4. Utilization of
NWM-based products in precise GNSS analyses, i.e. corresponding to the second
goal, is covered by Sect. 3.5. Various GNSS reprocessing and processing model
assessments are then summarized in Sect. 3.6. Finally, the transfer of knowledge, the
establishments of new ACs and integration of new networks is completed in Sect.
3.7. The definition of the SINEX_TRO V2 format, which was elaborated in a close
collaboration with WG3, is included in appendix D.
3.2
Campaigns for Development of Advanced Tropospheric
Products
The main goal of this section is to introduce two campaigns suitable for developing
and evaluating advanced tropospheric products. The first is the so-called
GNSS4SWEC WG1 Benchmark campaign which is considered as a cornerstone
that helped to accomplish various WG1 objectives within the COST Action. The
second is the Real-time demonstration campaign, which helped to develop, optimize
and evaluate new real-time GNSS software and products.
3.2.1
Benchmark Campaign – Common Data Set for New
Product Development and Validation1
M. Kačmařík
Institute of Geoinformatics, VŠB Technical University of Ostrava, Ostrava, Czech
Republic
e-mail: [email protected]
J. Douša
Geodetic Observatory Pecný, RIGTC, Ondřejov, Czech Republic
e-mail: [email protected]
G. Dick
GFZ German Research Centre for Geosciences, Helmholtz Centre Potsdam,
Potsdam, Germany
e-mail: [email protected]
1
Parts from this section were previously published in Douša et al. 2016
38
J. Douša et al.
F. Zus
GFZ German Research Centre for Geosciences, Potsdam, Germany
e-mail: zusfl[email protected]
R. Brožková
Czech Hydrometeorological Institute, Prague, Czech Republic
e-mail: [email protected]
H. Brenot
Royal Belgian Institute for Space Aeronomy, Uccle, Belgium
e-mail: [email protected]
A. Stoycheva
National Institute of Meteorology and Hydrology, Sofia, Bulgaria
e-mail: [email protected]
G. Möller
Department of Geodesy and Geoinformation, TU Wien, Wien, Austria
e-mail: [email protected]
J. Kaplon
Wrocław University of Environmental and Life Sciences, Wrocław, Poland
e-mail: [email protected]
Only basic information is presented here about the Benchmark campaign. For further
information please see Douša et al. (2016).
3.2.1.1
Motivation
An idea to create a well-prepared and extensive common data set which would
enable an effective collaboration within the WG1 itself, but also with other working
groups, arosed during the first WG1 meeting in Valencia, October, 2013. The natural
motivation for the campaign was to support the WG1 main goals, namely the
development of advanced GNSS tropospheric products in support of weather forecasting and vice versa exploitation of numerical weather data in precise GNSS
positioning. The Benchmark campaign planning started with inventory of requirements based on a wide discussion within the Action members. The following
requests for the Benchmark data set were summarized as follows:
• Period covering a month at least to enable NWM and GNSS processing initializations and to cover different weather conditions – quiet and variable, optimally
including a severe weather event.
• Availability of a dense network of GNSS reference stations in Europe with a
limited scale, but including flat and mountainous areas.
• Availability of meteorological data from independent sources.
3 Advanced GNSS Processing Techniques (Working Group 1)
3.2.1.2
39
Description of Selected Spatial and Temporal Domain
Finally, an area in central Europe was selected covering Germany, the Czech
Republic, Poland and Austria. Due to its size, the whole area was firstly divided
into a ‘core’ domain where below mentioned severe weather events took place and
an ‘extended’ domain which surrounded the ‘core’ one. In the second step both
domains were geographically divided into several clusters to allow a reasonable
GNSS data handling, see Fig. 3.1.
From the time perspective 2 months in June 2013 (May and June) were chosen.
The weather conditions in the selected area during May 2013 were mostly quiet. On
the contrary an extreme precipitation event lasting from May 31 to June 2 led to
devastating flooding on Danube, Elbe and Vltava rivers. Since the event was only
partly forecasted by NWM it was a suitable candidate for a GNSS meteorology
benchmark campaign as it could provide additional observations for meteorological
community. Significant precipitation periods hitting areas of smaller extent occurred
also from June 9 to June 11 and from June 23 to June 26.
3.2.1.3
Description of Collected Data Set
The Benchmark data set contains following data: GNSS observations and auxiliary
products, E-GVAP operational GNSS products, synoptic meteorological observations, NWM fields, radiosonde observations, WVR observations, meteorological
radar images. Collected data from individual sources are briefly described in paragraphs below. The data set is stored on an ftp server at Geodetic Observatory Pecný
Fig. 3.1 Benchmark core (yellow area) and extended domains depicted together with nine clusters
for GNSS stations (coloured points). The size of the points indicates height of GNSS reference
stations above the WGS-84 ellipsoid
40
J. Douša et al.
(GOP) and available to members of the COST ES1206 Action for research purposes.
To obtain more information and access to the data set please send an email to michal.
[email protected] or [email protected].
GNSS Observations observations from 430 GNSS reference stations in RINEX
format with 30s sampling interval were collected in total from which 247 sites
belonged to the ‘core’ domain. An average distance between two stations was
about 50–70 km. From the total number of GNSS sites, 4 observed GPS, GLONASS
and GALILEO satellites, 356 observed GPS and GLONASS satellites, and
remaining 70 stations were equipped with GPS recievers only. Station metadata
files were completed and checked carefully. A qualitative and quantitative control
and a standard positioning were performed using G-Nut/Anubis software
(Václavovic and Douša 2016). Besides multiple correction of metadata, 15 sites
had to be rejected from the data set because of the data quality issues.
E-GVAP Operational GNSS Products operational near real-time tropospheric solutions provided by 14 analysis centres for all GNSS reference stations in Europe were
collected for the campaign. These products contributed routinely to the E-GVAP
(http://egvap.dmi.dk) and are stored in the COST-716 format with a temporal
resolution of ZTD estimates from 5 to 60 min.
Synoptic Meteorological Data meteorological measurements of at least atmospheric air pressure, air temperature and relative humidity from 610 synoptic stations
were collected. Original data were provided in various formats which were, additionally, converted into a single unified plain text format with a sampling interval
ranging from 10 to 60 min.
NWM Data and Products NWM 3D data fields from the Czech Hydrometeorological Institute’s (CHMI) local area model ALADIN-CZ were extracted in GRIB
format. The horizontal resolution of the model is 4.7 4.7 km, outputs are provided
at 87 model levels with a 6-h interval of analysis run (00:00, 06:00, 12:00 and 18:00
UTC) and 1-h interval for forecast range. The model didn’t assimilate any GNSS
tropospheric products for delivered fields and provided necessary parameters for
derivation of hydrostatic and non-hydrostatic GNSS signal path delays as well as for
calculating the effect on signal due to so called hydrometeors (e.g. ice and liquid
water).
Radiosonde Data radio soundings from two different sources were collected providing profiles with full and reduced resolutions. Measurements with high resolution
were available from two sites in the Czech Republic – Prague-Libuš and Prostějov,
both provided by the CHMI. Altogether 278 files were obtained for the period of
Benchmark campaign. Radiosonde data with reduced vertical resolution from 19
European stations were provided by E-GVAP based on the EUMETNET – EUREF
MoU (Pottiaux et al. 2009).
Water Vapour Radiometer Data observations from two Water Vapour Radiometers
(WVR) situated in Germany were collected. The first one was operated by German
Research Centre for Geosciences (GFZ) in Potsdam (POTS) 30 km south-westward
3 Advanced GNSS Processing Techniques (Working Group 1)
41
from Berlin and provided measurements of IWV and liquid water in two modes:
slant (GPS satellite tracking) and zenith. The second was operated by Deutscher
Wetterdienst (DWD) at the Lindenberg meteorological observatory (LDBG) located
approximately 100 km eastward from Berlin and provided the same parameters as
the first mentioned WVR but only in the zenith direction.
Meteorological Radar Data raster images of combined observations from two
C-band Doppler meteorological radars located in the Czech Republic and operated
by the CHMI were collected. They represent a maximum reflectivity fields with side
projections in horizontal resolution of 1 1 km and 30-min time interval. The area
effectively covered by those two radars includes the territory of the Czech Republic
and areas of approximately 100 km outside the Czech state boundary.
Data Acknowledgement the members of COST Action ES 1206 thank all the
institutions that provided data for the campaign. GNSS data from the Austrian
network EPOSA were provided by Österreichische Bundesbahnen Infrastruktur
AG; GNSS data from SAPOS network in Germany by Zentrale Stelle SAPOS in
Hannover; GNSS data from several networks in the Czech Republic – (1) CZEPOS
by the Czech Land Survey Office, (2) Trimble VRS Now® by GEOTRONICS Praha,
s.r.o. and (3) GEONAS and VESOG stations thanks to the project CzechGeo
(LM2010008) operated by the Institute of Rock Structure and Mechanics of the
Academy of Sciences of the Czech Republic and the Research Institute of Geodesy,
Topography and Cartography, respectively; GNSS data from Polish ASG-EUPOS
network by the Head Office of Geodesy and Cartography in Poland; Synoptic data
by Zentralanstalt für Meteorologie und Geodynamik, ZAMG (Austria), Deutscher
Wetterdienst, DWD (Germany), Czech Hydrometeorological Institute, CHMI (the
Czech Republic) and Polish Institute of Meteorology and Water Management
(Poland). Finally, special thanks come to people assisting in collecting all the data
and metadata, namely: Dr. Jan Řezníček (GNSS/CZEPOS), Dr. Uwe FeldmannWestendorff and Dr. Markus Ramatschi (GNSS/SAPOS), Dr. Petr Novák (RADAR/
CHMI), Dr. Martin Motl (RAOBS/CHMI), Dr. Anna Valeriánová (SYNOP/CHMI),
Dr. Pavla Skřivánková (CHMI), Dr. Roland Potthast (SYNOP/DWD), Dr. Jürgen
Güldner (WVR-Lindenberg/DWD) and Dr. Stefan Heise (WVR-Potsdam/GFZ).
3.2.1.4
Reference Products and Their Initial Validation
After the data preparation, quality checking and cleaning, GNSS and NWM reference (tropospheric) products were generated and consequently initially validated.
Results of these steps provided a first insight into variations of parameters and
atmospheric conditions during the Benchmark campaign and were helpful for
more detailed planning of following Benchmark-related activities. Following paragraphs provide basic information about mentioned reference products and their
validation.
42
J. Douša et al.
GNSS Reference Tropospheric Products two institutions delivered their GNSS
tropospheric products for all GNSS reference stations within the Benchmark data
set. The first reference one was generated at GOP using the BSW52 (Dach et al.
2015) with the network processing approach using double-differenced GNSS
observations. The strategy for daily solutions was consistent with the GOP contribution to the EUREF Repro2 campaign (Douša et al. 2017). The second reference
tropospheric product was delivered by GFZ using the GFZ EPOS software (Gendt
et al. 2004; Ge et al. 2006) based on undifferenced GNSS observations and PPP
approach. GOP solution used CODE precise orbit and clock products while GFZ
solution was based on GFZ own precise products. The models were in case of both
GOP and GFZ solutions compliant with the IERS Conventions (Petit and Luzum
2010).
NWM Derived Tropospheric Products analogously to the situation with GNSS
reference products also NWM derived reference tropospheric products were independently generated by GOP and GFZ. GOP used their G-Nut/Shu software (Douša
and Eliaš 2014) and provided three products. Two were based on global numerical
weather models – the European Centre for Medium-Range Weather Forecasts
(ECMWF) ERA-Interim (Dee et al. 2011) and the Global Forecast System (GFS)
of the National Centres for Environmental Prediction (NCEP) available at http://
www.ftp.ncep.noaa.gov/data/nccf/com/gfs/prod/, and one was based on regional
model ALADIN-CZ which was provided within the Benchmark dataset. The following tropospheric and meteorological parameters were calculated for all stations:
zenith hydrostatic and wet delays, air pressure, partial water vapour pressure, mean
temperature, temperature lapse rate, water vapour pressure and zenith wet delay
exponential decay rates. GFZ used their direct numerical simulation (DNS) tool (Zus
et al. 2014) in order to derive the following parameters from two global NWMs
(ERA-Interim and GFS): zenith hydrostatic and zenith wet delays, horizontal (1st,
2nd order) tropospheric gradients and coefficients of hydrostatic and wet mapping
functions.
Validation of Reference Tropospheric Products Values of ZTD, ZWD and horizontal tropospheric gradients derived from described GNSS and NWM reference
products were compared to study the quality and mutual agreement. Table 3.1
summarizes comparison results of GNSS ZTDs with those derived from NWMs.
Mean statistics over all 430 sites demonstrated that both GNSS reference products
based on a completely different software and strategy performed very similarly when
compared to NWM products. Also, both software used for derivation of GNSS
related tropospheric parameters from NWM fields agreed very well. In case of
individual NWM the high-resolution ALADIN-CZ model outperformed both global
reanalysis models in the Benchmark domain and period mainly in terms of standard
deviation. For the NCEP’s GFS products a negative mean bias of about 5 mm was
observed compared to all other solutions. This bias stems from the bias in ZWD
values and its possible explanation is the low vertical resolution of GFS model which
resulted in larger interpolation errors. A comparable bias for GFS model was
reported by Urquhart et al. (2011).
NWM source
(software)
ERA (Shu)
ERA (Shu)
ERA (DNS)
ERA (DNS)
GFS (DNS)
GFS (DNS)
ALADIN (Shu)
ALADIN (Shu)
Grid
resolution
1 deg
1 deg
1 deg
1 deg
1 deg
1 deg
4.7 km
4.7 km
Analysis
[hour]
6
6
6
6
6
6
6
6
Forecast
[hour]
0
0
0
0
3
3
0,1,2,3,4,5
0,1,2,3,4,5
GNSS source
(software)
GOP (BSW52)
GFZ (EPOS-8)
GOP (BSW52)
GFZ (EPOS-8)
GOP (BSW52)
GFZ (EPOS-8)
GOP (BSW52)
GFZ (EPOS-8)
Pairs
#
224
224
224
224
224
223
1343
1343
Excl
#
2
3
3
3
7
7
20
22
Bias
[mm]
+0.0
+0.3
0.4
0.1
4.9
4.5
+0.8
+0.6
Sdev
[mm]
9.6
9.7
9.4
9.6
11.0
10.9
7.6
7.3
RMS
[mm]
10.0
10.0
9.8
9.8
12.0
11.8
7.8
7.5
3 Advanced GNSS Processing Techniques (Working Group 1)
Table 3.1 Comparison of zenith total delays from NWM and GNSS (mean values, 430 sites)
43
44
J. Douša et al.
Results of individual GNSS and NWM reference products were studied also
using maps showing differences between ZTDs from GNSS GOP product and
individual NWM products. Generally, a good homogeneity was observed from the
statistical results. Exceptions existed mainly in relation to the orography which
triggered larger differences between models particularly in mountain areas with
complex terrain where regional model ALADIN-CZ performed better than both
global models.
3.2.1.5
List of Benchmark Campaign Participants and Users
• G. Möller (TU Wien). In the first instance, we will use RINEX data to compute
STDs using different processing strategies. In addition, the GNSS GOP products
are used as input for GNSS tomography and assimilation studies.
• E. Pottiaux (ROB). The idea is to use the Benchmark campaign to assess the
different tropospheric products (RT, sub-hourly, hourly and post-processing –
ZTD, gradients and possibly slant delays from the BSW52) and refine my
processing strategies. By taking part myself in the Benchmark processing, I
also would like to stimulate the interaction with WG2 developments.
• J. Douša et al. (GOP). Generation of the reference GNSS tropospheric products.
Development of the software and the strategy for optimal provision of real-time
products and asymmetry monitoring, slant delay retrievals. Assessment of
NWM-derived tropospheric products and developing the combination of GNSS
and NWM data.
• G. Dick et al. (GFZ). Generation of the reference GNSS tropospheric products.
• M. Kačmařík (TU Ostrava). Inter-comparison of tropospheric slant delays form
GNSS, NWM and WVR at dual-station.
• W. Rohm et al. (WUELS). We would like to join the Benchmark campaign by
providing Slant Delays for selected stations (same as Michal’s) based on the
GNSS observations and ray-tracing.
• H. Brenot (BIRA). I would like to look at hydrometeors delays from ALADIN
outputs and to show their impact for the severe flood event of June.
• K. Eben (ICS ASCR). Assimilation of NRT ZTDs into the WRF mesoscale
model.
• T. Hadaś et al. (WUELS). The goal is to use the stored real-time products for
simulated real-time troposphere monitoring (ZTD estimates using original software), to optimize the methodology and algorithms, to compare results with other
real-time AC.
• L. Morel (Le CNAM). To process some stations of the benchmark campaign and
to deliver CNAM results (GAMIT processing).
• S. Nahmani (IGN). I want to verify some of my results on these specific stations.
• P. Gołaszewski (UWM). My research is focused on real time and post-processed
ZTD/ZWD estimation. Using this data will allow me to present the results on the
COST workshop in Potsdam, in September this year. I am interested in using
3 Advanced GNSS Processing Techniques (Working Group 1)
•
•
•
•
45
pseudo real-time demonstration (for ZTD estimation) and observation data for
post-processing.
S. de Haan (KNMI). My plan is to eventually assimilate the slant observations in
Harmonie. But first, I am going to compare the observation with my model
equivalent.
Arpaci (UBIMET). We will carry out case studies to examine the impact of the
assimilation on the forecast quality. Especially the catastrophic flooding from
June 2013, which affected big parts of central Europe seems to be an ideal
evaluation case study scenario. We will compute WRF model runs with ZTD
assimilation and compare them with runs without data assimilation. UBIMET
will carry out eye to eye verifications, Upper Air verifications (using sounding
and aircraft data) and surface verifications using VERA analysis data.
D. Kwasniak (UWM). I want to use them for my research about GPS positioning
using a new positioning method called MAFA method. Results of this research I
want to present on 17th Czech-Polish Workshop.
Y. Altiner et al. (BKG). Want to obtain the access for a processing the Benchmark
GNSS data in real-time simulated mode.
3.2.2
GNSS Real-Time PPP Demonstration Campaign
P. Václavovic
Geodetic Observatory Pecný, Research Institute of Geodesy, Topography and
Cartography, Zdiby, Czech Republic
e-mail: [email protected]
J. Douša
Geodetic Observatory Pecný, RIGTC, Ondřejov, Czech Republic
e-mail: [email protected]
F. N. Teferle
University of Luxembourg, Luxembourg, Luxembourg
e-mail: [email protected]
Providing new real-time or ultra-fast tropospheric products, such as ZTD, GRD,
STD, IWV maps or other derived products estimated using data from GNSS
permanent networks, is interesting for numerical and non-numerical weather
nowcasting and severe weather event monitoring (Guerova et al. 2016a, b). The
Precise Point Positioning (PPP) processing strategy plays a key role in the production of real-time tropospheric parameters because of its high processing efficiency,
and the sensitivity to the absolute value of the tropospheric delay. It enables to
exploit optimally data from all available GNSS multi-constellations, and facilitates
the production of all interesting GNSS parameters such as ZTDs, GRDs or STDs.
46
J. Douša et al.
Most importantly, the PPP is supported with the global orbit and clock products
provided by the RTS (Caissy et al. 2012) of IGS (Dow et al. 2009).
During the period 2015–2017, the COST Action ES1206 WG1 played an initiative role in the coordination of the development and the evaluation of GNSS realtime tropospheric products thanks to the design of the GNSS4SWEC Real-time
Demonstration Campaign which is briefly introduced in this subsection.
3.2.2.1
Campaign Design and Contribution Specifications
A list of common stations was selected from the E-GVAP supersites, EPN and IGS
sites, which were limited by a maximum of 50 in total, regionally and globally
distributed:
• E-GVAP super-sites (5): BRST, GOPE, ONSA, YEBE, ZIM2
• EPN sites (10): CASC, HERT, HOFN, MALL, MATE, NICO, PDEL, POTS,
REYK, WTZR
• IGS sites (17): ADIS, ALBH, ALGO, ALIC, AUCK, DUBO, LHAZ, NKLG,
NRMD, OHI3, POVE, THTI, ULAB, UNSA, WIND, YAR3, YELL
Station metadata are introduced using the RINEX skeleton files which are
available from EPN CB (http://www.epncb.oma.be/stations/log/skl) and IGS CB
(http://igscb.jpl.nasa.gov/igscb/station/general/skel). The strategy for tropospheric
estimation is generally free when optimized with respect to the individual software
capabilities and following the state-of-the-art models, in particular IERS conventions, antenna phase offsets and variations, a priori tropospheric model, and mapping
functions, and others.
Use of GNSS systems GPS and GPS + GLONASS is recommended if supported
by the software and real-time data streams, other systems are optional, however, the
constellation has to be properly defined in the header of the COST-716 file using the
PCD flag. The GPS and GPS + GLONASS solutions are provided in different files
and different names and individual solution (including variants) has to be specified
by the fourth character in the processing centre name. The character ‘G’ will be used
for GPS-only results, and the character ‘R’ will be used for GPS + GLONASS. In
order to support GPS + GLONASS, the IGS03 real-time product is mandatory for
the utilization to guarantee a consistency of mandatory products and enable to
compare the results. Thus even GPS-only solution should use the IGS03 stream,
for others solutions the precise ephemeris source is optional.
Parameters must be estimated with a 5-min resolution (parameter sampling rate).
If the higher sampling rate is used in the processing delivered product files should be
reduced to 5 min.
Parameters to be estimated: ZTD (mandatory) with the product sampling rate of
5 min (processing sampling rate can be higher), horizontal tropospheric gradients
(optionally) and coordinates (mandatory) estimated as static parameters. The contributors submit files with troposphere parameters every hour to the ftp-server at the
Geodetic Observatory Pecny after the registering of the product. The product file is
3 Advanced GNSS Processing Techniques (Working Group 1)
47
converted to the latest COST-716 format with the file name following the COST-716
conventions, i.e. using “demo” product status. Analysis centre providing more
product lines should be uploaded as separate COST-716 files when using a specific
analysis centre acronym (i.e. consisting of a unique analysis centre and a
product line).
3.2.2.2
Contribution and Monitoring
From April 2015 till the end of the COST Action, eight agencies succeeded to start
the real-time processing and provide partial contribution at least to the
GNSS4SWEC Real-Time Demonstration Campaign. Real-time/ultra-fast solutions
were provided using six different software and using various flavours of processing
options (Table 3.2). Truly real-time solutions using the operational processing
engine was provided by seven contributors.
For the purpose of a feedback to such product providers, a dedicated web service
for an easy monitoring and comparison of individual contributions Fig. 3.2. The
access has been made available also to a wide community (in particular interested
from the GNSS4SWEC WG2) at http://www.pecny.cz/COST/RT-TROPO to enable
visualising site-specific time series of recently estimated ZTD and gradient parameters from real-time solutions. For the comparison purpose, also near real-time
regional and global solutions from the GOP analysis centre operationally contributing to the EIG EUMETNET GNSS Water Vapour Programme, E-GVAP (http://
egvap.dmi.d) were included in the real-time demonstration monitoring.
Table 3.2 Contributions to GNSS4SWEC Real-Time Demonstration campaign
AC
GOP
Running agency
Geodetic Observatory Pecný, RIGTC
TUW
Technical University Vienna
ROB
Royal Observatory of Belgium
ASI
ULX
Agenzia Spaziale Italiana/Centro di
Geodesia Spaziale, Matera
University of Luxembourg
Software
G-Nut/
Tefnut
TUW
software
G-Nut/
Tefnut
GipsyOasis
BNC
TUO
Technical University of Ostrava
RTKLib
15.6.
2015
5.11.2015
BKG
Bundesamt für Kartographie und
Geodäsie
GFZ German Research Centre for
Geosciencies
BNC
1.3.2016
EPOSRT
16.2.1017
GFZ
Start
9.4. 2015
15.4.
2015
23.4.
2015
5.5. 2015
Update
Realtime
Realtime
Realtime
Hourly
Realtime
Realtime
Realtime
Realtime
Solutions
GPS, GLO,
gradients
GPS
GPS, GLO,
gradients
GPS,
gradients
GPS
GPS
GPS, GLO
GPS, GLO
48
J. Douša et al.
Fig. 3.2 Web service for the monitoring of the GNSS4SWEC Real-Time Demonstration campaign
3.2.2.3
Link to 4.3.7 IAG Working Group
The activity within the GNSS4SWEC project plays also a key role in the IAG
Working Group 4.3.7 ‘Real-Time troposphere monitoring’ which has been
established with the following objectives for the period of 2015–2019:
• Stimulate the development of software that enable routine production of realtime/ultra-fast tropospheric products.
• Develop optimal strategies suitable for numerical or non-numerical weather
nowcasting applications, and severe weather event monitoring.
• Demonstrate a reliable high-temporal resolution real-time/ultra-fast production,
assess applied method, software and precise real-time orbit and clock products.
• Evaluate real-time/ultra-fast tropospheric parameters and their potential for applications in meteorology.
• Setting up a link to the users, review product format and requirements.
3.3
Tropospheric Asymmetry Monitoring and Advantage
of Multi-GNSS
This section focuses on the detection of meteorological heterogeneities surrounding
ground-based GNSS stations, in particular the disturbance of GNSS signal through
the neutral atmosphere as retrieved by geodetic software with gradient and residual
contributions to Slant Total Delay (STD). The STD of the neutral atmosphere,
measured by GNSS technique, is the result of the adjustment of two components –
3 Advanced GNSS Processing Techniques (Working Group 1)
49
isotropic and anisotropic. The ZTD of the neutral atmosphere represents the isotropic
contribution above a GNSS site. To adjust the anisotropic contribution, the concept
of horizontal gradients has been introduced in GNSS software.
Additionally, this section included selected results of the multi-GNSS processing,
which is expected to foster and improve tropospheric parameters, in particularly a
possibility to monitor anisotrophy by estimating tropospheric horizontal gradients or
retrieving STDs from carrier-phase post-fit residuals for more satellites in view.
3.3.1
Concept of Tropospheric Gradients
H. Brenot
Royal Belgian Institute for Space Aeronomy, Uccle, Belgium
e-mail: [email protected]
The accuracy of slant delay measurements of neutral atmosphere (Latm) and the
number of visible satellites are critical for identifying the exact location of smallscale asymmetric tropospheric structures (Fig. 3.3).
STD ¼ Latm ðє; αÞ ¼ 10
6
Z
S
!
N P ds
ð3:1Þ
Latm is commonly called slant tropospheric delay (for the elevation є and the
azimuth α). Its formulation shows a dependency on the atmospheric refractivity (N )
!
and the position of source of signal ( P ). The adjustment of the slant path delay of
GNSS signal (Latm) through a blob of water vapour, can be well retrieved taking into
account the 1st order anisotropic character of the neutral atmosphere (model
presented Fig. 3.3; see Gradinarsky 2002).
Fig. 3.3 Illustration of
tropospheric heterogeneity
affecting path travel (S) of
GNSS signal. P⃗ is the
vector position along the
satellite direction (S) at the
elevation (є), and ρ is its
projection on the
horizontal plan
50
J. Douša et al.
Fig. 3.4 Example of the
distribution of the
atmospheric refractivity
(N) above a GNSS site for
which a model of a flat tilted
atmosphere is considered to
retrieve delay gradient. The
tilted line on the top,
corresponds to the isoline
for a constant refractivity N0
The atmospheric anisotropy reflects the fact that the speed of microwave signals
emitted from satellites and recorded by ground-based receivers differs according to
the azimuthal direction through the neutral atmosphere (and the asymmetrical
distribution of components surroundings a GNSS station). An inclined plane
model of troposphere (Fig. 3.4) schematised by linear thickness and density variations is considered to define horizontal gradients during the adjustments of tropospheric parameters (Davis et al. 1993; Gradinarsky 2002). The correction provided
by gradients possesses its own mapping function, mfaz (Chen and Herring 1997). The
expression of azimuthal anisotropic contribution (Laz) to the reconstruction of slant
delays depends on the satellite direction (elevation and azimuth).
According to Davis et al. (1993), the first order Taylor expansion (when ρ tends to
0⃗) of the atmospheric refractivity (N ) can be formulated
!
!
!
∂N ρ ; z
!
N ρ ; z ¼ N o ðzÞ þ
¼ N o þ ξ ðzÞ: ρ
!
!
!
∂ρ
ρ !0
ð3:2Þ
This expresses the 1st order anisotropy around a site. The assumption of a straight
line propagation is considered, with no time dependency of N (Gradinarsky 2002).
!
N0 is the isotropic contribution to N with cylindrical symmetry, and ξ is the
horizontal refractivity gradient. Both depend on the altitude z (function of the
elevation є). Using expression Eq. 3.2 injected in Eq. 3.1, the formulation of Latm
is obtained:
Latm ðє; αÞ ¼ 10
6
Z
S
N o ðzÞ ds þ 10
6
Z
¼ Lsym ðєÞ þ Laz ðє; αÞ
S
!
!
ξ ðzÞ: ρ ds
ð3:3Þ
Lsym is the symmetric contribution to ZTD mapped in direction of satellite, and
Laz is the asymmetric contribution to the total delay (with azimuthal dependency).
!
Vector position ρ in the horizontal plan can be expressed using the orthonormal
! !
base uNS ; uEW of the North-South (NS) and East-West (EW) directions:
3 Advanced GNSS Processing Techniques (Working Group 1)
!
ρ
51
!
!
¼ z: cot ðєÞ: cos ðαÞ uNS þ sin ðαÞ uEW
ð3:4Þ
A way to link the differential of length along the path of GNSS signal (ds) and the
differential of length along the vertical (dz) is to use mapping function (Niell 1996;
Böhm et al. 2006a, b) which can depend on the elevation (m0(є))
ds ¼ m0 ðєÞ : dz
ð3:5Þ
The refractivity gradient can be expressed in the base
!
!
!
uNS ;
!
!
uEW
:
ξ ðzÞ ¼ ξNS ðzÞ uNS þξEW ðzÞ uEW
ð3:6Þ
Considering Eqs. 3.3, 3.5 and 3.6, Laz can be formulated
Laz ðє; αÞ ¼ 10 6 : m0 ðєÞ: cot ðєÞ:ðGNS : cos ðαÞ þ GEW : sin ðαÞÞ
ð3:7Þ
!
(G , GEW) are the components of the horizontal delay gradients G in the base
!NS !
uNS ; uEW expressed as
!
!
!
G¼ GNS uNS þGEW uEW ¼
GNS
GEW
ð3:8Þ
The analytic formulation of (GNS, GEW) is following
GNS
GEW
0Z
1
1
z:ξ
ð
z
Þ
dz
NS
B
C
C
Z0
¼B
@ 1
A
z:ξEW ðzÞ dz
ð3:9Þ
0
The unit of delay gradient is the same as for the delay – the unit of length [m]. The
!
delay gradient G can be defined as a correction of phase residual projections
depending on the elevation and azimuth angles of visible satellites. An interpretation
of the horizontal gradient is for example that, a gradient component of 1 mm at the
zenith will show a higher correction for elevation of 45 and 25 (correction of about
2 mm and 4 mm, respectively).
Following the implementation of tropospheric delay (ZTD), the horizontal delay
gradient is the second tropospheric parameter implemented in the least-squares
adjustment proceeded by geodetic software in the analysis of GNSS data (Davis
et al. 1993; MacMillan 1995; Alber et al. 1997; Chen and Herring 1997; Bar-Sever
and Kroger 1998). Initially, horizontal gradients were introduced into the calculations in order to improve positioning solutions, showing a 15% improvement for the
horizontal repeatability (Bar-Sever and Kroger 1998) and a 25% improvement
52
J. Douša et al.
between wet delays from GNSS and WVR. Afterwards, the potential of delay
gradient for GNSS meteorology has been investigated. Walpersdorf et al. (2001)
have showed how the use of GNSS gradient can describe the approach of a front
towards Marseille in the south-east of France in 1998. Iwabuchi et al. (2003) also
showed that the temporal and spatial variations of GNSS gradients matched well
with the moisture field determined by ZTD and with the meteorological condition in
summer 1996 over the Japan Islands (in particular during the passage of a weather
front).
The interest of using GNSS gradient for monitoring severe weather is still
investigated, especially to improve our understanding of meteorological situation.
For example, during the flash-flood event of September 2002 (southeastern France),
three phases have been identified by Delrieu et al. (2005) and confirmed by GNSS
data (Brenot et al. 2006). The maximum daily precipitation reached 691 mm.
Figure 3.5 shows the path of a MCS over CHRN station during the Phase III
(from 01:00 to 18:00 UTC on 9 September 2002). It can be seen that horizontal
gradient can clearly be used to monitor tropospheric structure (MCS or blob of water
vapour). Just after this rainfall event (in Autumn 2002), a dense network of 20 GPS
stations has been installed by the Mediterranean Hydrometeorological Observatory
Cévennes-Vivarais (OHM-CV) to proceed monitoring of convective systems and
Cévenol effect, eastwards of Mont Aigoual (AIGO station located at 30 km of this
dense network; see Fig. 3.5). Using different configurations in calculations (settings,
geometry, constrains), the precision of NS and EW delay gradients components
(zenith direction) has been evaluated to 0.35–0.7 mm and 0.2–0.5 mm respectively
(Brenot et al. 2014a, b). The lack of GPS satellites at low elevation in the North for a
Fig. 3.5 (left) Time-series of delay gradients over Château-Renard (CHRN) GPS station during the
flash-flood event of September 2002 (over Bouche-du-Rhône, between Montpellier and Marseille,
France). Purple double-arrow shows the path of a quasi-stationary Mesoscale Convective System
(MCS) over CHRN station. (right) GNSS delay gradients are superposed over radar reflectivity
localising the MCS close to MTPL, CHRN and VERC stations. Dash circles show representativity
areas of gradients (for a cutoff angle of 10 ).
3 Advanced GNSS Processing Techniques (Working Group 1)
53
network located at a latitude of 45 N, explains why the precision of the NS
component is less good than the EW one.
3.3.2
Global Validity and Behaviour of Tropospheric
Gradients Estimated by GPS2
L. Morel
École Supérieure des Géomètres et Topographes, Le Mans, France
e-mail: [email protected]
Estimation of tropospheric gradients in GNSS data processing is a well-known
technique to improve positioning. Today, they are routinely estimated by several
global and regional GNSS analysis centres but they are still not yet used for
operational meteorology. We have studied the physical meaning of tropospheric
gradients estimated from GPS observations recorded by several permanent stations
located all around the world. In a first study with several stations on Corsica island,
we estimated ZTD and tropospheric gradients using two software: GAMIT/GLOBK
(GAMIT version 10.5) and GIPSY-OASIS II version 6.3 in order to analyse the
differences in the tropospheric results (ZWD and gradients) coming from the
processing strategy (double-differences for GAMIT/Globk versus zero-difference
for GIPSY-OASIS). That study allowed to confirm a strong correlation between the
two software for ZWD estimation (98%) and a good correlation for gradient
estimation (70%). No direct correlation with elevation or geographical location has
been noticed but the gradients were oriented inward land (Fig. 3.6), in opposite
direction from tropospheric humidity field processed by ERA – Interim and with a
direction relatively stable along the year (Morel et al. 2014).
In a following study with 14 stations all around the world, selected due to their
proximity of the relief, we also observed that gradient directions were stable over the
time and pointed toward the relief for most of the stations selected. Correlation
coefficients were processed between gradients (yearly mean values (Ge, Gn) as
vector component) and direction of the steep slopes (obtained by analysing Digital
Elevation Model at 20 km, 40 km and 60 km around the station), see Table 3.3.
These results gave us a first step for a physical meaning to gradients when stations
are close to high mountains. We can notice 10 stations with a correlation coefficient > 0.4 (60 km) and 2 stations without any correlation (BOGT and CHWK) but
surrounded by mountains. Now, we are going to continue the study with more
stations and years and quantify multipath effect.
2
Parts from this section were previously published in Morel et al. 2014
54
J. Douša et al.
Fig. 3.6 Monthly mean
gradient magnitude and
direction (blue arrows
estimated by GAMIT and
red arrows estimated by
GIPSY-OASIS) for all
stations in Corsica Island
over the year 2011. Gradient
vectors are drawn
considering monthly mean
values (Ge, Gn) as vector
components
Table 3.3 Correlation coefficient between gradient and relief around 14 stations (red: negative
correlation, orange: weak correlation and green: strong correlation)
3 Advanced GNSS Processing Techniques (Working Group 1)
3.3.3
55
Monitoring of Severe Weather from Wet Gradients,
Residuals and Slants3
H. Brenot
Royal Belgian Institute for Space Aeronomy, Uccle, Belgium
e-mail: [email protected]
To retrieve Slant Wet delays (SWD) in direction of GNSS satellites (for elevation є
and azimuth α), two contributions are commonly considered (see Eq. 3.10): the
wet
isotropic contributions (Lsym
) with spherical symmetry and derived from Zenith Wet
wet
Delays (ZWD), and the anisotropic contributions (Lasym
) with azimuthal asymmetry
and derived from horizontal wet gradients and residuals.
wet
wet
SWD ¼ Lsym
ðєÞ þ Lasym
ðє; αÞ
ð3:10Þ
wet
wet
Lsym
ðєÞ ¼ ZWDadjusted mf sym
ðєÞ
ð3:11Þ
with
The GNSS technique retrieves Zenith Tropospheric Delay (ZTD) of the neutral
atmosphere using an a priori ZHD (ZHDapriori) and adjusting a ZWD
(ZWDadjusted ¼ ZTD – ZHDapriori). ZHDapriori is generally obtained using the
formula of Saastamoinen (1972); see also Davis et al. (1985) and Elgered et al.
wet
depends on the elevation (є) of each satellite and using a wet mapping
(1991). Lsym
function, e.g. GMF (Boehm et al. 2006a, b).
wet
Concerning the wet anisotropic contribution (Lasym
), the 1st and the 2nd order
wet
contributions can be considered by using respectively the wet gradients (Laz
) and
the one-way post-fit residuals (Lres), as formulated in Eq. 3.12. Generally, the
contribution of residuals is not considered because it can be highly affected by
multipath and artefacts in calculations. This study tries to show you the interest of
using residuals, as retrieved by GAMIT software (Herring et al. 2010). The 1st order
wet
wet
wet
contribution to Laz
is estimated with gradient wet components (GNS
, GEW
) that is
connected
to
the
azimuth
(α),
and
with
the
use
of
gradient
mapping
function
wet
mf az ¼ 1=ð sin єtanє þ CÞ which depends on satellite’s elevation and on a
constant C (Chen and Herring 1997).
wet
wet
Lasym
ðє; αÞ ¼ Laz
ðє; αÞ þ Lres ðє; αÞ
with
3
Parts from this section were previously published in Brenot et al. (2013)
ð3:12Þ
56
J. Douša et al.
wet
wet
wet
wet
Laz
ðє; αÞ ¼ mf az
ðє; CÞ: GNS
: cos ðαÞ þ GEW
: sin ðαÞ
The gradient components (GNS, GEW) retrieved by GNSS technique are total. This
means there is no distinction between wet and hydrostatic gradients (Chen
and Herring 1997; Flores et al. 2000). In geodetic software it is commonly considered that C ¼ 0.0032 (Herring 1992), but for the estimation of the asymmetric
wet
wet delay (Laz
), C ¼ 0.0031 can be used (Chen and Herring 1997). The wet gradient
!
wet
G
is expressed by the difference of the hydrostatic to the total component.
!
wet
G
¼
wet
GNS
wet
GEW
¼
GNS
GEW
Ghydrostatic
NS
Ghydrostatic
EW
!
ð3:13Þ
To obtain the hydrostatic gradient components, a characterisation of the surface
pressure field around each GNSS station is required. In that case, the hydrostatic
gradient can be established by fitting a plane through the pressure measurements
(Champollion et al. 2004; Brenot et al. 2014a, b). From the pressure field near a
GNSS site, the spatial variations of the hydrostatic delay per unit of distance (km) in
the north-south (Z hydrostatic
) and east-west (Z hydrostatic
) directions can be calculated.
EW
NS
Generally, for a case study, surface pressure measurements around all GNSS stations
are not available. Outputs from numerical weather model can be considered. Assuming an exponential law in the hydrostatic refractivity and considering the scale height
of the gradients in the hydrostatic delays set to H ¼ 13 km (as suggested by Chen and
Herring 1997), the spatial variations of the hydrostatic delay can be converted in
hydrostatic gradients (Elósegui et al. 1999; Ruffini et al. 1999; Flores et al. 2000) to
obtain wet gradient components (Eq. 3.14).
Gwet
NS
Gwet
EW
¼
GNS
GEW
Z hydrostatic
NS
H:
Z hydrostatic
EW
!
ð3:14Þ
The 2nd order asymmetric contribution from residuals (Lres), can be considered if
a station is weakly affected by multipath or these avoided by Multipath Stacking
Methods (MPS), as introduced by Elósegui et al. (1995) and Shoji et al. (2004). By
using MPS, low elevation measurements can be improved by identifying and
avoiding non-tropospheric signature in data. Estimates of slant delays can also be
improved by using Phase Centre Variation model (PCV) for the antenna (Shoji et al.
2004). This aspect is not treated in this study, as well as the questioning of artefacts
in calculations. This study investigates the contribution of residuals (Lres) to wet
delays without any correction of multipath and PCV model.
Brenot et al. (2013) have studied in detail the rainfall event of 28–29 June 2005.
This paper shows the critical role of GNSS horizontal gradients of the water vapour
content to detect small scale structures of the troposphere (i.e. convective cells), and
presents a strategy to identify typical water vapour configurations (dry/wet dipole in
3 Advanced GNSS Processing Techniques (Working Group 1)
57
Fig. 3.7 (a) Imaging of the 2D field of ZWD with a classic interpolation (stations are plotted using
!
black circles); (b) improvement of this field by GNSS gradients. Wet gradients Gwet are plotted
using grey arrows at each GNSS site. BUGG, ERPE, GERA and BRUS stations are plotted.
Locations of 9 major Belgium cities (red circles) and meteorological radars (yellow triangles) are
also plotted on these 2D maps
time and space) and obtain preliminary signs of the initiation of deep convection.
The complementary objective of this work is to investigate, step by step, the use of
wet gradient to improve 2D field of ZWD and visualise water vapour blobs, and the
use of SWD delays to monitor small scale tropospheric structure of convective cells.
Total GNSS gradients has been used by Brenot et al. (2013, 2014a, b) to improve
the spatial resolution of the 2D field of ZTD (using Hermite interpolation or pseudoobservation defined by gradients). The strategy of Brenot et al. (2013) with additional pseudo-observations has been transferred to the improvement of the 2D field
!
of ZWD with wet gradients Gwet (comparison between classic interpolation and
improved ZWD field is shown in Fig. 3.7). The grey arrows in Fig. 3.7b, represent
!
wet
G (expressed in the zenith direction) with amplitudes of about 0.01 m and more
for stations GERA and BUGG at 12:00 UTC on 29 June 2005. A relevant way to
visualise, in time, isotropic and anisotropic contributions to wet delays, is proposed
in Fig. 3.8. Such a graph acts by superimposing horizontal wet gradient to ZWD.
Four stations have been selected (ERPE, BUGG, GERA and BRUS), as being
several time close or overflown by convective cells. Let’s focus our attention on
the time window (10:00 to 14:00 UTC) when these four stations measure high
anisotropic contributions (wet gradient higher than 0.01 m).
The morning of the 29th June 2005, the wet delay is gradually decreasing for the
four stations (as shown in Fig. 3.8), then a sudden increase is observed (starting at
11:00 UTC for ERPE and GERA, and 12:00 UTC for BRUS and BUGG). Figure 3.9
shows the spatial distribution of wet delay all over Belgium at 12:15, 12:30, 13:00
and 13:15 UTC. Low ZWD is observed by BUGG station at 12:15 in Fig. 3.9a. At
this moment, the initiation of convection is taking place between BUGG and BRUS
(dry/wet dipole). At 12:30 UTC (in Fig. 3.9b), a strong contrast of moistening is still
58
J. Douša et al.
!
Fig. 3.8 Visualisation of isotropic (ZWD) and anisotropic (wet gradient, Gwet ) contributions to
delays on 29 June 2005 for 4 stations (a) ERPE, (b) BUGG, (c) GERA, and (d) BRUS
observed between BRUS and BUGG with significant rainfall. Wet gradients of
BUGG and BRUS stations point to area with high precipitation.
At 13:00 UTC, the flux of water vapour from north to south is separated by a drier
area on the north and east side of BRUS, where the initiation of deep convection is
taking place (on the east side of BRUS). The strength of the rapid flux of moistening
is shown by a strong increase of the field of ZWD close to BUGG and ERPE and in
the north-west of Belgium (see Fig. 3.9c). Wet gradient amplitudes higher than
0.015 m have been observed for several stations during this rainfall event (and
especially during this time window), notably at NAMR station (south-east of
BRUS in Fig. 3.9d); see Brenot et al. (2013). Looking at the improved 2D field in
Fig. 3.9, dry/wet contrast of ZWD field is a good indicator of preliminary signs of
deep convection and heavy precipitation.
An investigation of the interest of SWD for monitoring small-scale structures
(sub-kilometric size) for these four stations is presented in Fig. 3.10 (time window
from 10:00 to 14:00 UTC). Using skyplots, the three contributions to SWD (isotrowet
pic contribution by ZWD and Lsym
, and anisotropic contributions by wet gradients
wet
and residuals, respectively Laz and Lres) are shown for GNSS signals from two
satellites recorded by four stations.
Small wet gradients and residuals are observed between 10:00 and 12:00 UTC for
these four stations (except a wet structure at 11:00 on the west side of ERPE and
3 Advanced GNSS Processing Techniques (Working Group 1)
!
59
Fig. 3.9 (left) 2D fields of ZWD improved by, Gwet at (a) 12:15, (b) 12:30, (c) 13:00 and (d) 13:15
UTC on 29 June 2005; (right) Radar precipitation at (a) 12:15, (b) 12:30, (c) 13:00 and (d) 13:15
UTC on 29 June 2005.
60
Fig. 3.10 Skyplots for
4 GNSS stations of couples
SWD/satellites trajectories
on 29 June 2005, (a) ERPE,
(b) BUGG, (c) GERA, and
(d) BRUS
J. Douša et al.
3 Advanced GNSS Processing Techniques (Working Group 1)
61
BUGG, 15 min later, specifically where the initiation of a convective system took
place). Between 12:00 and 14:00, several tropospheric structures are identified by
the four stations, showing a base of anisotropic contribution provided by the wet
gradient and a precise time-space detection by the residuals.
Even similarity and correlation with radar precipitation is not obvious, neither
straightforward. It can be noticed that at 12:15, a blob of water vapour is detected on
the west side of ERPE. For BUGG station, the negative residuals of PRN27 satellite
at 12:15, decrease the anisotropy seen by the wet gradients (showing that the highest
density of humidity of the wet structure is not located at 60 of elevation). No
significant structure is seen by Lres for GERA station. However, a tropospheric
structure is detected by residuals of PRN27 BRUS station (in the south – southeast direction for elevations between 40 and 50 ).
At 12:30, the wet gradients of BRUS indicate the north, and Lres of PRN27 is
negative for elevation between 50 and 70 , showing a drier structure (convective
system has moved eastward). A wet structure is seen by Lres of PRN27 for ERPE
station (located north-west of BRUS).
Between 13:00 and 13:30, PRN27 residuals of ERPE and BRUS stations indicate
wet structure in the east direction for an elevation of 70 . BUGG residuals show a
wet structure in the south-east.
There is a clear identification of sub-kilometric meteorological structure by
one-way post fit residuals. Even no correction of multipath and PCV model has
been applied (Elósegui et al. 1995; Shoji et al. 2004), a way to justify that the
structures detected by residuals have properly a tropospheric origin is to compare
Lres on 29 June with the one measured the day before, for which the trajectories of
satellites are very similar. On 28 June, the tropospheric activity was moderate.
Figure 3.11 shows clearly low residuals on 28 June, justifying that the structures
detected on 29 June are due to tropospheric activity (no multipath or artefact effect).
If a GNSS station is a good candidate (confirmation of the tropospheric origin of
residuals), the 2nd order asymmetric contribution from residuals (Lres), can be
considered in meteorological applications (assimilation in numerical weather forecasts or imaging for nowcasting). This study shows a good potential for residuals and
SWD for detecting small scale tropospheric structures affecting signal propagation
between GNSS satellites and stations.
3.3.4
Indicator of Tropospheric Activity Based
on the Disruption of GNSS Signals4
H. Brenot
Royal Belgian Institute for Space Aeronomy, Uccle, Belgium
e-mail: [email protected]
4
Parts from this section were previously published in Brenot and Warnant (2008)
62
Fig. 3.11 Skyplots for
4 GNSS stations of one-way
post fit residuals (Lres) and
satellites trajectories on
28–29 June 2005, (a) ERPE,
(b) BUGG, (c) GERA,
and (d) BRUS
J. Douša et al.
3 Advanced GNSS Processing Techniques (Working Group 1)
63
The aim of this subsection is to find a new indicator of small-scale tropospheric
activity. Different candidates of tropospheric effects indicators can be considered
according to GNSS carrier phase measurements (King et al. 1985; Dong and Bock
1989; Blewitt 1989; Leick 1989; Teunissen et al. 1998). However, ZTD and
gradients are not the best candidate being the results of a time and space average.
STDs show a good potential to detect small-scale meteorological structure (see Sect.
3.3.5). Nevertheless, double differences (L1, L2) of the ionosphere-free combination
of GNSS phase observations can be used as an additional detection of the presence of
small-scale structures in the troposphere. Small-scale structures induce disturbances
on phase measurements. This study considers stations for which the positions and
ij
the geometric distancesDAB
are precisely known between couples of satellites and
ij
ij
and the ambiguity N AB
ground-based receivers. The tropospheric perturbation T AB
, IF
remains the only unknown parameters in the double difference of phase of the
ij
ionosphere-free (IF) combination ϕAB
, IF (for a simplified mathematical model of
phase measurements (Seeber 2003; Leick 2004; Brenot and Warnant 2008). A new
ij
observable of phase ΦAB
, IF can be estimated:
ij
ij
ΦAB
, IF ¼ ϕAB, IF
f 1 ij
f ij
ij
D ¼ 1 T AB
þ N AB
, IF
c AB
c
ð3:15Þ
c is the speed of electromagnetic waves, ( f1, f2) carrier frequencies of (L1 or L2).
ij
The ambiguity term (N AB
, IF ) has the following expression:
i
ij
N AB
, IF ¼ N A, IF
N Bi , IF
N Aj, IF
N Bj, IF
ð3:16Þ
Ambiguities ( N Ai , IF , N Bi , IF , N Aj, IF and N Bj, IF ) are defined using ionosphere-free
ij
combination. The phase ambiguity term (N AB
, IF ) is a real number with a constant
value.
ij
Figure 3.12 shows an example of tropospheric perturbation T AB
and ambiguity
ij
ij
N AB, IF presented by the phase observable ΦAB, IF (called IF Double Difference and
expressed in cycles) on 29 June 2005 for BRUS-GILL baseline (4 km) and the
couple of satellites (27-08). Without the presence of the troposphere, a constant real
ij
value of the IF DD should be observed according to the ambiguityN AB
, IF (which can
be a real number). The error induced by the troposphere on the IF Double Difference
ij
observable ΦAB
, IF time-series is clearly shown in Fig. 3.12 between 12:00 and 13:00
UTC. According to radar imaging important precipitations (higher than 100 mm/h)
took place over and north-east of OLLN station at 12:30 UTC the 29th June 2005
(location of strong tropospheric activity).
A high content of water vapour and the existence of hydrometeors induces a
strong perturbation of atmospheric refractivity (Brenot et al. 2006). Perturbation of
refractivity can clearly explain sudden variability of tropospheric error T Ai measured
by station A for a signal emitted by a satellite i. The following expression presents
64
J. Douša et al.
Fig. 3.12 IF Double Difference of BRUS-GILL baseline the 29th June 2005 (Day Of Year 180)
the relation of tropospheric error T Ai (i. e. generally called STD) with neutral
atmosphere refractivity (N ):
T Ai
¼ 10
Z
6
ð3:17Þ
N ds
ds is a differential distance according to path travel of signal between satellite
ij
i and station A. The tropospheric error T AB
induces the perturbation of the phase
ij
observable ΦAB, IF , as defined by Eq. 3.15, representing the double difference of
phase of the ionosphere-free combination with a correction of the geometric distances. This tropospheric error has the following expression:
ij
T AB
¼ T Ai
T Bi
T Aj
T Bj
ð3:18Þ
Sudden perturbations of tropospheric errors (T Ai ,T Bi ,T Aj and T Bj ) by a small-scale
ij
ij
structures induce direct perturbations of T AB
and ΦAB
, IF . Considering two epochs of
measurements (epoch t0 and epoch t0 + Δt, for example Δt ¼ 5 min), Fig. 3.13
illustrates the direct impact of the occurrence of a small-scale tropospheric structure
ij
ij
on phase measurements (observables ΦAB
, IF and T AB ).
The resolution of the ambiguities is required. For a selected Day Of Year (DOY),
reference satellites are chosen to form double differences (DD) and maximise the
time periods. The atmospheric scans by these couples of satellites are sufficient to
represent the tropospheric activity. Considering NAMR-OLLN baseline DOY
180 of 2005 (couple of satellites 10–21) and BRUS-BERT baseline DOY 365 of
3 Advanced GNSS Processing Techniques (Working Group 1)
65
Fig. 3.13 Perturbation of TABij induced by a small-scale tropospheric structure for two epochs of
measurements (epoch t ¼ t0 and epoch t ¼ t0 + Δt)
Fig. 3.14 IF Double Difference of NAMR-OLLN baseline the 29th of June 2005 event (DOY 180)
on the left, BRUS-BERT baseline on the right (no meteorological event on 31 December 2005,
DOY 365). The fits of DD time-series with polynomial functions of the 3rd order are shown
2006 (couple of satellites 16–19), Fig. 3.14 shows IF Double Differences time-series
ij
(observable of phase ΦAB
, IF of Eq, 3.15) for these two baselines (called IF DD plotted
with crosses).
The impact on DD depends on the elevation of considered satellites. This is a
specificity of the tropospheric activity. In order to display only the influence of
small-scale structures on DD time-series, fits of IF DD time-series have been
assessed using polynomial functions of the 3rd order (dashed line Fig. 3.14) and
66
J. Douša et al.
Fig. 3.15 1 h radar precipitation accumulation starting at (a) 16:00 UTC, (b) 17:00 UTC, (c) 18:00
UTC, and (d) 19:00 UTC on 2006/10/01
Fig. 3.16 Radar precipitation on 29 June 2005 at 14:30 UTC (left), and on 31 December 2006 at
13:15 UTC (right)
the biases between IF DD and the respective fits, called IF DD Residuals, can be
obtained.
The estimation of the bias to the fit removes elevation effects. Then small-scale
structures are clearly identified for NAMR-OLLN baseline on the 29th of June 2005
(in Fig. 3.14 between 14:00 and 15:00 UTC). Figure 3.15 shows a time-series of IF
DD Index. To obtain this Index of the tropospheric activity, absolute values of IF DD
Residuals (in cycles) is converted into centimetres (multiplying by the wave length:
19.029 cm).
According to radar imaging of rain rate (in Fig. 3.16 on the left), the tropospheric
small-scale activity around Namur (NAMR and OLLN stations) during DOY 180 of
2005 can be easily observed between 14:00 and 15:00 UTC. Note that a strong
tropospheric activity was also observed between 12:00 and 13:00 UTC this day (see
Figs. 3.14 and 3.15; see also Brenot et al. 2013). No tropospheric activity took place
around Brussels (station BRUS) DOY 365 of 2006 at 13:15 UTC (see radar imaging
in Fig. 3.16 on the right).
3 Advanced GNSS Processing Techniques (Working Group 1)
67
Fig. 3.17 Imaging of maximal IF DD Index detected (a) 29 June 2005 at 14:30 UTC; (b)
31 December 2006 at 13:15 UTC
Considering all the couples of satellites for a selected baseline and all the
available phase measurements, the daily tropospheric activity (superposition of all
the IF DD Index of the selected satellites-stations couples) can be shown (see Brenot
and Warnant 2008).
Considering every baseline of the Belgian network, IF DD Index imaging are
shown Fig. 3.17. In this imaging, geometric segments (each one corresponding to a
baseline) are affected by the maximum IF DD Index estimated at a given moment
(at 14:30 UTC on 2005/06/29 and 13:15 UTC on 2006/12/31, for the two examples
presented) according to all the couples of satellites considered in our system.
Note that the rainfall cell present over OLLN station does not appear with ZTD
imaging due to the time and space average. Horizontal delay gradient points a
direction where the local anisotropy is maximal. However gradient represents a
time and space average which punctually (at 12:30 UTC) do not show exactly the
location of small-scale structure (in the north-east direction of OLLN station), as
seen by IF DD Index.
The IF DD Index imaging (Fig. 3.17a) is clearly sensitive to sudden perturbation
of tropospheric activity. That means sensitive to the occurrence of tropospheric
small-scale structures which locally affect transmission of signals from GNSS
satellites at a given epoch and not for a time and space average measurements. IF
DD Index shows strong perturbations of GNSS signal propagation induced by the
troposphere around OLLN station between 14:30 and 15:00 UTC (Brenot and
Warnant 2008). The presence of water vapour and hydrometeors above OLLN and
on the north-east side of this station, affects Double-Difference observations for
OLLN-NAMR baseline the 29th June of 2005 (DOY 180). The deep convection
process and the thermally driven turbulent mixing that moves air parcels from the
lower to the upper atmosphere, shows a vertical extension up to 14 km close to
68
J. Douša et al.
BRUS, MECH and BUGG stations, with occurrence of heavy rain and, for some
area, hail stones at 12:20 UTC (Brenot et al. 2013).
Using a dense network of GNSS stations (e.g. the Belgian dense network with
baselines from 5 to 30 km), a relevant monitoring of tropospheric structure can be
established with IF DD Index. As an example of severe weather, the month of
September and the start of autumn 2006 was exceptionally hot and dry. During the
last days of September, sea breeze was finally bringing humidity in the warm low
layer. At higher altitude, a strong dynamic was taking place with strong jets
maintaining powerful forcing able to generate the sturdiest thunderstorms. The
differences of wind direction from low to high levels, so called wind shear, in
association with strong flux of water vapour and moistening of low level, led to a
critical situation on 1st October 2006. The tropospheric activity took place during all
this day, initiating locally deep convection and generating specific conditions for the
establishment of supercells that can stay active during few hours.
One of these supercells has created a convenient meteorological situation for the
formation of a tornado (see photography in Fig. 3.18). This supercell (associated
with heavy rainfall from a cumulonimbus cloud, as seen on radar imaging on the
south of Brussels in Fig. 3.19), has generated this tornado close to BRUS and LEEU
stations. Taking the northeast direction after its creation and avoiding densely built
area, the tornado nevertheless reached farms located along its trajectory. Several
buildings have been seriously damaged. Figure 3.18b presents the probability of hail
(maximum values) during the 1st October 2006 event (DOY 274). The passage of
the supercell on the south side of Brussels can be observed. Several other supercells
are also shown (wind direction oriented from south-west to north-east). Operational
hail detection products are derived from the height of the freezing level and from
45 dBZ echotop values provided by single-polarization C-band weather radar
(Delobbe and Holleman 2006). The supercell close to Brussels (BRUS station)
and Sint-Pieters-Leeuw (LEEU station) can clearly be observed by daily IF DD
Index (Fig. 3.19) applied for a baseline of 8 km.
100
90
80
70
60
50
40
30
20
10
0
Fig. 3.18 Photography of the tornado of Petit-Roeulx-lez-Braine (source: C. De Keyser) at 16:00
UTC on 2006/10/01 (left); daily probability of hail (in %) for this day, with positions and names of
6 GNSS stations; courtesy of Laurent Delobbe (right)
3 Advanced GNSS Processing Techniques (Working Group 1)
69
Fig. 3.19 Radar imaging on the 1st of October 2006 at 16:25 UTC (left); daily IF DD Index for
BRUS-LEEU baseline (right)
Note that the tropospheric daily IF DD Index detects strong activity DOY 274 for
baseline ANTW-KALL. On the other hand, a quiet tropospheric activity is observed
for the baseline VOER-TONG.
The interest of looking at radar imaging of hail probability is that the possible
production of hail requires an important vertical extension and a consequent amount
of water vapour, linked with the existence of hydrometeors. The IF DD Index at
16:25 UTC in Fig. 3.19 shows high level of activity with Index more than 10 cm.
In Fig. 3.20, 1 h radar precipitation accumulation shows that supercells have
traveled from south-west to north-east. Baselines BUGG-NIKL, BREC-HERE,
DIES-MOL0, LEEU-NIVL, BRUS-NIVL, BERT-OLLN and BREE-MAAS are
plotted on these radar imaging. For these baselines, IF DD Index tropospheric
activities are presented in Fig. 3.21.
According to radar precipitation accumulation imaging (Fig. 3.20), the description of Fig. 3.21 (left) is the following: at about 16:00 UTC, tropospheric activity for
baseline BUGG-NIKL (up to 7 cm) was occurring close to BUGG station (cell C of
Fig. 3.20a). This same cell C was located then close to HERE station at about 17:00
UTC and induced an IF DD Index up to 7 cm for BREC-HERE baseline. Around
18:00 UTC, the cell B was approaching over BUGG station (in Fig. 3.20b) and
inducing IF DD Index up to 10 cm for BUGG-NIKL baseline. The passage of the
cell B can clearly be observed with IF DD Index presented in Fig. 3.21 between
18:00 and 19:30 UTC. Around 19:00 UTC, the cell B (in Fig. 3.20c) was above
HERE station (IF DD Index up to 11 cm for BREC-HERE baseline), and at 19:20
UTC (in Fig. 3.20d) above MOL0 station (IF DD Index up to 8 cm for DIES-MOL0
baseline).
According to radar precipitation accumulation imaging (in Fig. 3.20), the description of Fig. 3.21 (right) is the following: at 16:20 UTC, tropospheric activity has
taken place around LEEU and NIVL stations (IF DD Index up to 8 cm for baseline
LEEU-NIVL). Between 16:30 and 17:00 UTC the supercell A has moved from
south-west to east of Brussels (successively IF DD Index of 6 cm for BRUS-NIVL
and BERT-OLLN baselines). Around 18:40 UTC, the cell A was close to MAAS
station and induced IF DD Index up to 7 cm for BREE-MAAS baseline. The passage
of the supercell A is clearly shown from 16:00 to 19:00 UTC on IF DD Index.
70
J. Douša et al.
Fig. 3.20 1 h radar precipitation accumulation starting at (a) 16:00 UTC, (b) 17:00 UTC, (c) 18:00
UTC, and (d) 19:00 UTC on 2006/10/01
The contribution of hydrometeors in association with water vapour bubble to
strong IF DD Index of tropospheric activity is indisputable. The blobs of water
vapour surround rainfall cells with a high vertical extension (up to 10 km for cells B
and C and up to 11 for cell A, as estimated by the maximum radar reflectivity from
the weather radar of Wideumont).
The implementation of this index started in the frame of the GALOCAD/ESA
project (2006–2008). The aim was to find relevant tropospheric and ionospheric
indicators to warn the impact on NRT positioning solutions, i.e. effect in
RTK-architecture for a dense network of stations (Brenot and Warnant 2008;
Warnant et al. 2008, Wautelet et al. 2008; Brenot et al. 2014a, b).
To summarise, this study presents GNSS indicators of meteorological activity
that allow the detection of small-scale structures in the neutral atmosphere. The
scope is to present a new NRT index of meteorological activity based on doubledifference of the ionosphere-free combination (so called IF DD index). Contrary to
3 Advanced GNSS Processing Techniques (Working Group 1)
71
Fig. 3.21 Daily IF DD index (2006/10/01) for BUGG-NIKL, BREC-HERE and DIES-MOL0
baselines (left), and for LEEU-NIVL, BRUS-NIVL, BERT-OLLN and BREE-MAAS baselines
(right)
ZTD imaging and horizontal delay gradients measurements from geodetic software
(result of a mean time and space solution), the IF DD Index imaging is clearly
sensitive to sudden disturbances of tropospheric activity. That means sensitive to the
occurrence of tropospheric small-scale structures which locally affect couples of
satellites emitted signals considered in NRT applications (i.e. GNSS meteorology or
positioning) at a given epoch. The use of the IF DD Index can be planed operationally in NRT meteorological or geodetic system using dense networks, being useful
for forecasters and nowcasting.
The next step of this work can be to improve the time and space imaging of the IF
DD Index using multi-GNSS satellites, in collaboration with forecasters. The contribution of hydrometeors to IF DD Index need to be investigated studying correlation with radar reflectivity. The use of the geometry-free combination (GF) can also
be used to define an indicator of ionospheric activity (GF DD Index). This work
based on DD difference can also be transferred to of L1, L2 and IF combination to
obtain IF Index of the tropospheric activity between a satellite and a ground-based
station. As a first investigation the flash-flood event in the Gard region, on 8–9 September 2002, has been tested (Brenot et al. 2006), showing in Fig. 3.22 strong
tropospheric activity when convective cells were located close to CHRN station.
3.3.5
Validation of Slant Tropospheric Delays5
M. Kačmařík
Institute of Geoinformatics, VŠB Technical University of Ostrava, Ostrava, Czech
Republic
e-mail: [email protected]
5
Parts from this section were previously published in Kačmařík et al. (2017)
72
J. Douša et al.
Fig. 3.22 Time-series of IF combination (GPS satellite PRN06) and its fit for CHRN station on
8–9 September 2002 (left); IF Index for all the signals emitted from satellites and recorded by
CHRN on 8–9 September 2002 (right)
J. Douša
Geodetic Observatory Pecný, RIGTC, Ondřejov, Czech Republic
e-mail: [email protected]
P. Václavovic
Geodetic Observatory Pecný, Research Institute of Geodesy, Topography and
Cartography, Zdiby, Czech Republic
e-mail: [email protected]
G. Dick
GFZ German Research Centre for Geosciences, Helmholtz Centre Potsdam,
Potsdam, Germany
e-mail: [email protected]
F. Zus
GFZ German Research Centre for Geosciences, Potsdam, Germany
e-mail: zusfl[email protected]
H. Brenot
Royal Belgian Institute for Space Aeronomy, Uccle, Belgium
e-mail: [email protected]
G. Möller
Department of Geodesy and Geoinformation, TU Wien, Wien, Austria
e-mail: [email protected]
E. Pottiaux
Royal Observatory of Belgium, Brussels, Belgium
e-mail: [email protected]
P. Hordyniec
Wrocław University of Environmental and Life Sciences, Wrocław, Poland
e-mail: [email protected]
3 Advanced GNSS Processing Techniques (Working Group 1)
73
J. Kaplon
Wrocław University of Environmental and Life Sciences, Wrocław, Poland
e-mail: [email protected]
L. Morel
École Supérieure des Géomètres et Topographes, Le Mans, France
e-mail: [email protected]
Tropospheric STD represents the total delay that undergoes the GNSS radio-signal
due to the neutral atmosphere along the path from a satellite to a ground receiver
antenna. It is evident that STDs can provide much more information about the
distribution of water vapor in the troposphere than classical ZTDs. With a continuous development of NWM forecasting and nowcasting tools a demand for highquality humidity observations with high spatial and high temporal resolutions is
growing at side of a meteorological community. On the other hand, despite a set of
studies which dealt with this topic there is still not a uniform consensus on how to
reconstruct STDs from GNSS processing results. Therefore, we decided to realize an
extensive inter-technique validation of STDs using data from Benchmark data set
and try to answer at least some of the opened questions. A summary of obtained
results is given here, however, we refer the reader to the publication of Kačmařík
et al. (2017) for a much more detailed presentation.
3.3.5.1
Description of STD Validation Study
From the complete Benchmark data set, we selected a subset of 10 GNSS reference
stations situated at six different locations (Table 3.4). It also includes collocated
(dual) GNSS stations playing an important role in the validation since they track
GNSS satellites with the same azimuth and elevation angles, so that they should
deliver the same or very similar tropospheric parameters used for STD
reconstructions.
Seven institutions delivered their STD solutions for this validation study, namely
Ecole Supérieure des Géomètres et Topographes (ESGT CNAM), Geodetic Observatory Pecný (GOP, RIGTC), Helmholtz Centre Potsdam – German Research
Centre for Geosciences (GFZ), Royal Observatory of Belgium (ROB), VŠB-Technical University of Ostrava (TUO), Vienna University of Technology (TUW), and
Wroclaw University of Environmental and Life Sciences (WUELS). Principal
information about individual solutions are given in Table 3.5.
In total, we validated eleven solutions computed with five different GNSS
processing software. Considering all available GNSS solutions, only GOP used a
stochastic modelling approach to estimate all parameters. Additionally, GOP provided two solutions: (1) GOP_F using Kalman filter (forward filter only), i.e. capable
of providing ZTD, tropospheric gradients and STDs in real time, and (2) GOP_S
applying the backward smoothing algorithm (Václavovic and Douša 2015) on top of
the Kalman filter. The latter improves the quality of all estimated parameters during
74
Table 3.4 Characteristics of 10 GNSS reference stations
Name
GOPE
KIBG
LDB0
LDB2
POTM
POTS
SAAL
WTZR
WTZS
WTZZ
Latitude [ ]
49.914
47.449
52.210
52.209
52.379
52.379
47.426
49.144
49.145
49.144
Longitude [ ]
14.786
12.309
14.118
14.121
13.066
13.066
12.832
12.879
12.895
12.879
Height [m]
593
877
160
160
145
144
796
666
663
666
Network
IGS, EPN
Dual station
IGS, EPN
LDB2
LDB0
POTS
POTM
IGS, EPN
IGS
IGS
WTZS, WTZZ
WTZR, WTZZ
WTZR, WTZS
Receiver
TPS NET-G3
TPS GB-1000
JAVAD TRE_G2T
JPS LEGACY
JAVAD TRE_G3TH
JAVAD TRE_G3TH DELTA
TPS GB-1000
LEICA GRX1200 + GNSS
SEPT POLARX2
JAVAD TRE_G3TH DELTA
Antenna
TPSCR.G3 TPSH
TPSCR3_GGD CONE
JAV_GRANT-G3T NONE
LEIAR25.R4 LEIT
JAV_GRANT-G3T NONE
JAV_RINGANT_G3T NONE
TPSCR3_GGD CONE
LEIAR25.R3 LEIT
LEIAR25.R3 LEIT
LEIAR25.R3 LEIT
J. Douša et al.
Solution Name
CNAM
GFZ
GOP_F
GOP_S
ROB_G
ROB_V
TUO_R
TUO_G
TUW_3
TUW_7
WUE
Institution
ESGT CNAM
GFZ Potsdam
GO Pecný
GO Pecný
ROB
ROB
TU Ostrava
TU Ostrava
TU Vienna
TU Vienna
WUELS
Strategy
DD
PPP
PPP
PPP
DD
DD
DD
DD
PPP
PPP
PPP
Software
GAMIT
EPOS 8
G-Nut/Tefnut
G-Nut/Tefnut
BSW52
BSW52
BSW52
BSW52
NAPEOS
NAPEOS
BSW52
GNSS
GPS
GPS
GPS
GPS
GPS + GLO
GPS + GLO
GPS + GLO
GPS
GPS + GLO
GPS + GLO
GPS
Elev. cut-off
3
7
7
7
3
3
3
3
3
7
3
Mapping function
VMF1
GMF
GMF
GMF
GMF
VMF1
VMF1
VMF1
GMF
GMF
VMF1
ZTD/gradients interval
1 h/1 h
15 min/1 h
2.5 min/2.5 min
2.5 min/2.5 min
15 min/1 h
15 min/1 h
1 h/3 h
1 h/3 h
30 min/1 h
30 min/1 h
2.5 min/1 h
Post-fit residuals
NO
YES
YES
YES
YES
YES
NO
NO
YES
YES
YES
3 Advanced GNSS Processing Techniques (Working Group 1)
Table 3.5 Information about GNSS-based STD solutions used in the validation
75
76
J. Douša et al.
the batch processing interval and eliminates effects such as the PPP convergence or
re-convergence. Some institutions delivered also two STD solutions which differ in
a single processing setting. The aim was to evaluate their impact on STDs: a)
TUO_G and TUO_R exploit GPS-only and GPS + GLONASS observations respectively, b) TUW_3 and TUW_7 apply an elevation cut-off angle of 3 and 7 degrees
respectively, and c) ROB_G and ROB_V use the GMF and VMF1 mapping
functions respectively. Additionally, ROB solutions are the only ones based on the
processing of double-difference (DD) observations and providing ZD carrier-phase
post-fit residuals converted from the original DD residuals using the technique
described in Alber et al. (2000).
For an independent validation of STDs from GNSS processing we used STDs
derived from NWM via ray-tracing and from observations of Water Vapor Radiometer (WVR). In case of NWM derived STDs, four institutions delivered their
solutions based on three different NWMs: ALADIN-CZ (4.7 km resolution,
limited-area hydrostatic model, operational analysis in 6-h interval with forecasts
for 0, 1, 2, 3, 5, 6 h, http://www.umr-cnrm.fr/aladin/), ERA-Interim (1 horizontal
resolution, 6-h reanalysis), and NCEP-GFS (1 horizontal resolution, 6-h operational
analysis,
https://www.ncdc.noaa.gov/data-access/model-data/model-datasets/
global-forcast-system-gfs). None of these NWM assimilates data from ground
GNSS stations. For more information about the models, see Douša et al. (2016)
and specifically Trojáková (2016) for the ALADIN-CZ model and Dee et al. (2011)
for the ERA-Interim reanalysis. First, STD solutions using the ERA-Interim and
NCEP-GFS models were delivered by GFZ Potsdam using acronym ERA/GFZ and
GFS/GFZ, respectively. Two STD solutions were then delivered for the ALADINCZ model: (a) ALA/BIRA generated at Royal Belgian Institute for Space Aeronomy
(BIRA), and (b) ALA/WUELS delivered by Wroclaw University of Environmental
and Life Sciences. For a description of these solutions we refer to Kačmařík
et al. (2017).
In case of WVR we used the instrument operated at GFZ Potsdam. The instrument is situated on the same roof as the GNSS reference stations POTM and POTS.
The WVR is switching between ‘zenith mode’ when it is measuring IWV and ‘slant
mode’ when it is tracking GPS satellites using an in-built GPS receiver. In the latter
case, Slant Integrated Water Vapour (SIWV) values are delivered for the direction of
satellites. Our study focuses on the comparison of STDs, not SIWV. It was thus
necessary to convert the WVR SIWV into STDs. The used conversion of WVR
SIWVs to STDs aimed at minimum distorting the accuracy of original WVR
observations and is described in Kačmařík et al. (2017).
3.3.5.2
Introduction to STD Estimation from GNSS Observations
The tropospheric STD is being reconstructed from tropospheric parameters valid in
zenith direction which are being estimated during GNSS observation processing. It
can be expressed by Eq. 3.19, where ZHD and ZWD represent Zenith Hydrostatic
3 Advanced GNSS Processing Techniques (Working Group 1)
77
and Wet delay, respectively, G horizontal tropospheric gradient and ele is the
elevation angle and azi is the azimuth angle of observation.
STDðele; aziÞ ¼ ZHD ∙ mf h ðeleÞ þ ZWD ∙ mf w ðeleÞ þ Gðele; aziÞ þ RES
MPT
ð3:19Þ
The elevation angle dependency of STD is described by the mapping functions,
separately for the hydrostatic (mfh) and the wet (mfw) components. Additionally,
post-fit residuals (RES) may contain un-modelled tropospheric effects not covered
by the estimated tropospheric parameters. Obviously, residuals contain also other
un-modelled effects such as multipath (MPT), however, potentially including also
errors from antenna phase centre variations or systematics in satellite clocks. For
eliminating such systematic effects, cleaning of post-fit residuals was applied by
generating elevation/azimuth-dependent correction maps as described by Shoji et al.
(2004). We thus computed mean values of post-fit residuals in 1 1 degree bins
using the whole benchmark period for each solution and station. Computed means
were then subtracted from the original post-fit residuals to generate solutions using
cleaned residuals. Therefore, whenever zero-differenced (ZD) post-fit residuals were
available for any solution delivered to the validation, three variants of the solution
were used: (1) solution without residuals (nonRES), (2) solution with raw residuals
(rawRES), and (3) solution with cleaned residuals (clnRES).
Example maps obtained with gradient estimation, polar 1 1 degree bins for
multipath determination and 2 sigma outlier rejection threshold are presented on
Fig. 3.23 for LDB0. Kapłon et al. (2017) later realized a set of tests to evaluate the
impact of strategy of STDs calculation on STD differences obtained from GNSS and
Multipath map for station: LDB0
0
40
20
–50
–150
–100
–50
0
50
100
150
Map of std. dev. of multipath: LDB0
50
80
40
60
30
40
20
20
10
0
–150
–100
–50
0
50
Azimuth [deg]
100
Fig. 3.23 Example equal area maps of multipath effect for LDB0 station
150
StdDev. [mm]
Elevation angle [deg]
60
Multipath [mm]
50
80
78
J. Douša et al.
raytracing through the GFS NWM model including testing the impact of method of
multipath effect calculation: polar degree bins (1 1, 2 2 and 5 5 degree) or
equal area bins (Huisman et al. 2009) of 1, 2 or 5 degree height, level of outliers
reduction (2 or 3 sigma) on STDs. Summarizing results of these tests, all 24 variants
of multipath maps provided very similar results with a slight edge for variants using
the smallest 1 1 degree bins.
3.3.5.3
Methodology of STD Comparisons
Since NWM outputs are restricted to the time resolution of their predictions (typically 1, 3 or 6 h) and since WVR is able to track only one satellite at one moment, all
three sources provide different numbers of STDs per day. Therefore, three different
comparisons were realized: (1) results for GNSS versus GNSS comparisons,
(2) results for GNSS versus NWM comparisons, and (3) results for GNSS versus
WVR comparisons. All the presented results were obtained over the whole benchmark period of 56 days. No outlier detection and removal procedure was applied
during the statistics computation within the study. Two variants of the comparisons
are presented: ‘ZENITH’ and ‘SLANT’. ‘ZENITH’ stands for original STDs
mapped back to zenith direction using 1/sin(ele) formula. Such mapping aimed at
normalizing STD differences for their evaluation in a single unit. The ‘SLANT’ type
of comparison denotes an evaluation of STDs at their actual elevation angles. To be
more specific, STDs were grouped into individual elevation bins of 5 degrees, i.e. for
example all STDs with an elevation angle between 10 and 15 degrees were evaluated
as a single unit. The cut-off angle of 7 degrees was used in all GNSS versus GNSS
and GNSS versus NWM comparisons. In GNSS versus WVR comparisons
15 degrees cut-off was applied to exclude problematic WVR observations from
low elevation angles.
3.3.5.4
GNSS Versus GNSS: Evaluation of All GNSS Solutions Versus
the Reference GNSS Solution
Individual GNSS solutions were first compared to the GFZ solution in the zenith
direction (ZENITH). Figure 3.24 shows all the solutions using STDs calculated from
the estimated ZTD and horizontal gradient parameters, i.e. without adding post-fit
residuals. Adding raw or clean residuals, applied consistently to both compared and
reference solutions, provided very similar graphs (not displayed). Colours in the
Figure indicate the processing software used in individual solutions. Medians of all
solutions (dotted lines in each bin) are displayed for each station in order to highlight
differences among the stations. These were observed mainly as biases ranging from
3.6 mm to 0.6 mm. The better agreement between GOP and GFZ solutions could
be attributed to a similar strategy of both solutions compared to others. It is
particularly visible for LDB0 and POTM stations where median values over all
solutions differ by 2.3 mm and
3.6 mm, respectively. The reason for the
3 Advanced GNSS Processing Techniques (Working Group 1)
79
Fig. 3.24 Comparison of individual GNSS STD solutions against GFZ solution, all without using
residuals (nonRES) and projected in the zenith direction: bias (left) and standard deviation (right).
The median value of all solutions at each station is represented by the dotted blue line in each bin
divergent behaviour at the two stations has not been identified although site metadata
were cross-checked carefully. A significant difference can also be noticed for
TUW_3 and TUW_7 at the station KIBG where these solutions used individual
antenna calibration files while all others solution used type mean calibration (Schmid
et al. 2016). Plots with standard deviations show agreements within 3–5 mm among
all the stations and all solutions. The only exception is the GOP_F solution
representing a simulated real-time analysis applying only a Kalman filter (not
backward smoothing) and providing results by a factor of 2 worse compared to the
others in terms of precision.
3.3.5.5
GNSS Versus GNSS: Evaluation in the Slant Direction
Figure 3.25 provides an evaluation of the STDs at their original elevation angles for
the station POTS. Four individual panels show bias (top left), normalized bias
(NBIAS, top right), standard deviation (bottom left), and normalized standard
deviation (NSDEV, bottom right). Normalized bias and normalized standard deviation were computed to see the dependence of relative errors in STDs at different
elevations. For its computation, absolute differences of STDs from two solutions
were divided by the STD values from the reference solution.
We found that the agreement among individual solutions compared to the GFZ
STDs is rather stable above the elevation angle of 30 degrees. Corresponding biases
of individual elevation bins are within 4 mm and standard deviations are slowly
increasing up to 10 mm at 30 degrees. With elevation angles decreasing below
30 degrees the biases slightly increase for some solutions. Normalized standard
deviation remains almost constant over all elevation angles indicating a very consistent relative performance of STDs among all the solutions. A similar behaviour is
present at all stations although the absolute values can be higher for some stations or
solutions, namely GOP_F for LDB0 and WTZZ with standard deviations reaching
up to 72 mm.
80
J. Douša et al.
Fig. 3.25 Comparison of individual GNSS STD solutions against GFZ STD solution at station
POTS, in slant directions
3.3.5.6
GNSS Versus NWM: Summary of Results
A summary of the GNSS versus NWM validation is presented in Table 3.6. For each
reference station a median of bias and a median of standard deviation in the zenith
direction between all GNSS solutions and a particular NWM-based solution are
given. If we consider ALA/BIRA and ERA/GFZ only, without the two mountainous
stations KIBG and SAAL, absolute biases between NWM and GNSS solutions stay
mostly below 3 mm, which represents a very good agreement between these
independent sources used for retrieving slant delays. Standard deviations generally
range from 8 mm to 12 mm in the zenith projection, with an exception of
ALA/WUELS showing lower precision by a factor of 2.5.
3.3.5.7
GNSS Versus WVR: Summary of Results
A bias of about 5.5 mm in the zenith direction was found between WVR and GNSS
solutions at station POTS while the bias at station POTM was around 10 mm. The
difference between stations POTM and POTS are probably related to issues with
GNSS data processing at POTM. The bias between POTS and WVR roughly
corresponds to 1 kg/m2 of IWV, what can be addressed as the achievable accuracy
3 Advanced GNSS Processing Techniques (Working Group 1)
81
Table 3.6 Medians of bias and standard deviation values of differences between all GNSS
solutions and a particular NWM-based solution at each reference station, expressed in the zenith
direction
Station
GOPE
KIBG
LDB0
LDB2
POTM
POTS
SAAL
WTZR
WTZS
WTZZ
Bias (mm)
ERA/
ALA/
GFZ
BIRA
0.3
3.3
19.3
4.9
2.0
0.7
1.6
0.9
3.4
6.3
1.7
1.4
19.4
7.8
4.8
1.5
3.5
0.9
2.1
0.9
GFS/
GFZ
8.6
9.6
5.5
6.1
12.5
7.6
11.7
4.9
4.2
6.0
ALA/
WUELS
11.5
22.5
10.6
15.1
18.9
12.5
24.3
10.2
10.8
11.6
Standard deviation (mm)
GFS/
ERA/
ALA/
GFZ
GFZ
BIRA
8.3
10.3
7.1
11.6
17.8
11.0
9.9
10.3
8.5
9.1
10.1
8.6
8.0
10.6
9.4
7.7
10.3
9.2
12.7
17.9
11.8
11.0
11.8
8.5
11.4
12.3
8.7
11.3
12.0
8.9
ALA/
WUELS
22.4
26.7
26.2
25.4
26.2
25.8
22.9
23.1
23.7
23.7
of any technique, however, WVR accuracy is more dependent on a proper instrument calibration. Values of standard deviation around 12 mm in the zenith direction,
were higher than those observed in GNSS versus GNSS comparisons and slightly
higher than from GNSS versus NWM comparisons.
3.3.5.8
Results at Collocated Stations
For the GNSS versus NWM and GNSS versus WVR comparisons at individual
stations slightly higher values of standard deviations were always found for GNSS
solutions applying raw or cleaned residuals in contrast to versions of solutions
without any residuals (Kačmařík et al. 2017). However, since two erroneous techniques were always confronted to each other without knowing the true reference,
these results do not tell anything about potential of post-fit residuals.
For these reasons, we assessed all GNSS solutions at the collocated (dual) stations
because for them we are able to provide troposphere-free differences of STDs to
evaluate noise of GNSS STD retrievals. Dual stations were available in the benchmark campaign at three different locations in Germany. The first two sites collocate
twin GNSS reference stations (LDB0 + LDB2 and POTM+POTS), the third location
collocate three individual reference stations (WTZR+WTZS+WTZZ).
During normal weather conditions, the tropospheric variation is reasonably
smooth, meaning it can be well represented by GNSS STDs reconstructed only
from ZTDs and horizontal gradients. However, during high temporal or spatial
variabilities in the troposphere, post-fit residuals certainly contain tropospheric
signals which are not modelled. If they surpass the observation noise and other
residual errors from GNSS models, cleaned residuals should be considered in the
GNSS STD model as described in Eq. 3.19. In order to initially address optimal STD
modelling under different weather conditions within the benchmark, we tried to
82
J. Douša et al.
identify days with a high variability in the troposphere. Daily standard deviations of
cleaned post-fit residuals were computed individually for each day of the benchmark, for every station and GNSS solution for 1-degree elevation angle bins. We
studied their daily variations considering the GNSS model applied. If cleaned postfit residuals consist of the noise of observations only, the variation in time should be
negligible. However, the days showing significantly higher values, correlated at
collocated stations, indicated highly variable tropospheric conditions.
Three such days were identified at LDB0, LDB2, POTM and POTS stations (May
31, June 20, June 23) and 2 days at WTZR and WTZS stations (June 19, June 20).
They all very well correspond to the days initiating heavy precipitations in the
domain, Douša et al. (2016). Typical differences between raw and clean residuals
are displayed in Fig. 3.26 for all elevations during the normal day (June 19, DOY
170) and the following day with high variability in the troposphere (DOY 171, June
20) for POTM and POTS stations using GFZ solution. Obviously, the variability of
clean residuals (black dots) and their 2-sigma envelops are higher by a factor of two
for the day of year 171 compared to 170. The variability is clearly visible over all
elevations, but the increase is slightly higher at low elevations. The plot demonstrates the different quality of GNSS observations, particularly related to a multipath
effect displayed by 2-sigma envelop (green curves). The multipath level is much
lower for station POTS which is using a choke ring antenna compared to station
POTM which is not using a choke ring. The similar situation was found for stations
LDB0 and LDB2. Variability of 2-sigma envelopes of clean residuals (red curves)
indicates a higher sensitivity of clean residuals to the weather conditions compared
to station selection and observation quality, thus suggesting a significant
Fig. 3.26 Comparison of individual GNSS STD solutions against GFZ STD solution at station
POTS, in slant directions. Elevation-dependent variability of clean residuals (black dots) and their
2-sigma envelops (red curves) are showed for June 19 (DOY 171) and June 20 (DOY 170) and
stations POTS and POTM. Additionally, plots display 2-sigma envelopes for raw residuals (blue
curves) and multipath (green curves)
3 Advanced GNSS Processing Techniques (Working Group 1)
83
contribution from the troposphere to the cleaned residuals. In the same context, raw
residuals show much higher sensitivity to the observation quality compared to
different weather conditions, which is particularly true in case of LDB0 and LDB2
stations.
In a next step elevation-dependent differences of STDs from all three versions
(without residuals, with raw residuals, with clean residuals) were analysed for days
with high and low variability of residuals. We noticed following:
(a) STD differences are more or less similar for both days, i.e. no significantly
different between days with normal and high variations in the troposphere. It
suggests that increased residuals contain strong contributions from the tropospheric effect that could not have been assimilated into ZTDs and tropospheric
horizontal gradients
(b) STD differences using raw residuals were always the largest ones and they
varied with the elevation angle,
(c) relative performance of differences from STDs with clean residuals and without
residuals for different days remained similar. Uncertainties of the simplified
STDs at low elevations surpassed additional uncertainties due to applying
clean residuals. According to the magnitude of clean residuals at low elevations
(Fig. 3.26), the small uncertainties from calculated differences indicated the
presence of tropospheric signals in the residuals at low elevations, roughly
below 30 degrees. It seemed to be almost independent from the weather conditions and is supposed to represent mainly unmodelled horizontal asymmetry in
the troposphere.
Figure 3.27 displays results for comparisons of individual collocated stations in
slant directions calculated from all days of the benchmark. The same statistics and
plots (not displayed) were prepared also for days identified with ‘severe’ weather
conditions, but only minor differences were observed. Strong variations are
observed mainly in normalized biases over all elevation angles for the solutions
using raw post-fit residuals (rawRES) regardless weather conditions. These are
clearly related to local effects such as multipath or modelling instrumented related
effects (phase centre offsets and variations) and disappear after using the cleaned
residuals (clnRES). The standard deviations and normalized standard deviations at
all stations are clearly the lowest for variants without using post-fit residuals
(nonRES), slightly higher using cleaned residuals, and significantly higher when
using raw residuals, i.e. corresponding to above performed inter-technique
validations.
3.3.5.9
Future Work
Three institutions (GFZ, GOP, ROB) delivered GNSS STD solutions not only for
the ten GNSS reference stations but for the whole GNSS network within the
Benchmark data set. Our future study will therefore focus on (a) larger comparison
within the network of stations, (b) an evaluation of azimuthal dependency of post-fit
84
J. Douša et al.
Fig. 3.27 Comparison of GNSS STDs at dual stations computed over whole benchmark period
from individual GNSS solutions in the slant direction for dual stations from left to right: LDB0LDB2, POTM-POTS, WTZR-WTZS. Statistical parameters from top to bottom: bias, normalized
bias, standard deviation, normalized standard deviation
residuals under severe weather conditions and (c) an evaluation of GNSS STDs
estimated from real-time and post-processed solutions using a stochastic approach.
3.3.6
Information Content in Post-fit Residuals, PPP vs DD
Approach6
G. Möller
Department of Geodesy and Geoinformation, TU Wien, Wien, Austria
e-mail: [email protected]
Based on real GNSS measurements, the tropospheric signal in post-fit residuals is
difficult to assess since it is superimposed by a series of other unmodelled effects like
6
Parts from this section were previously published in Möller (2017)
3 Advanced GNSS Processing Techniques (Working Group 1)
85
observation noise, multipath, satellite clock or orbit errors. Nevertheless, observation
stacking methods as applied in (Kačmařík et al. 2017; Möller 2017) allow for the
reduction of common parts like multipath or clock errors, but only when longer time
periods or larger GNSS networks are processed. Hence, within the COST action an
initiative was carried out which addresses the general tropospheric signal in GNSS
post-fit residuals, with focus on differenced data processing.
Theoretically, the precise point positioning (PPP) and the double difference
approach (DD) are equivalent with respect to redundancy and with respect to the
estimates, in case a correct stochastic model is introduced. Practically, the DD
approach has some advantages in data processing since the satellite and receiver
clock errors and therewith the hardware biases cancel out, which allows the fixing of
integer ambiguities. Further, also the pre-processing is less critical since the receiver
clock error has to be known only with μs-accuracy.
Unfortunately the greatest strength of double-difference processing, the elimination of common effects, is also a shortcoming at the same time in small networks (<
500 km). In such networks, tropospheric parameters cannot be estimated in an
absolute sense but rather with respect to a reference station. Therefore, reference
values (station coordinates and ZTD) have to be introduced, at least for one station,
and constrained to their given values. Then the tropospheric parameters can be
estimated like in PPP processing, except for the reference station. In order to analyse
satellite or station specific effects in double-difference residuals (DDR), the residuals
have to be converted into zero-difference residuals (ZDR), also known as pseudoZDR since certain conditions have to be applied for the reconstruction. (Alber et al.
2000) suggested a two-step approach in which the DDR vector is converted into a
pseudo-ZDR vector, assuming zero-mean conditions.
In order to analyse the applicability of this approach and in general of the
tropospheric signal in DDR, two sets of dual-frequency GPS observations were
simulated for 12 stations in Austria. Both sets differ only with respect to the applied
troposphere model. While no troposphere model was applied to the first set of
observations, ZTDs and East-West gradients were simulated for the second set.
The observations of all 12 stations were processed in PPP and double-difference
approach. If both, ZTD and gradients are estimated, the simulated STD could be
recovered with sub-mm accuracy and the post-fit residuals became negligible. If only
the ZTD is estimated, it is expected that an anisotropic delay remains in the post-fit
residuals. It turned out that in case of PPP methods the anisotropic delay, except for a
small offset which was absorbed by the ambiguity parameter, could be recovered
from the PPP post-fit residuals but unfortunately not from the DDR. If only a single
baseline is processed, the DDR and also the reconstructed pseudo-ZDR are almost
zero since the anisotropic effects were differenced out in data processing. In best
case, the resulting STD bias and standard deviation was 1 mm +/ 37 mm. This
was obtained by fixing the ZTD and by taking all possible baselines between the
12 stations into account for the reconstruction of pseudo-ZDR using the method
proposed by (Alber et al. 2000). However, a comparable result (0 mm +/ 38 mm) is
obtained if no residuals are added to the isotropic STD.
86
J. Douša et al.
In practice, the ZTD is not known and therewith cannot be fixed to its given value.
Thus an additional solution was created whereby ZTD but no gradient parameters
were estimated. This results in an increase of bias and standard deviation (84 mm +/
105 mm). This example underscores the importance to estimate gradient parameters in addition to ZTDs. An unmodelled east-west gradient of 2 mm introduced a
ZTD error of 35 mm +/ 13 mm. For more details the reader is referred to (Möller
2017).
It becomes obvious that the applied reconstruction method proposed by (Alber
et al. 2000) is less suited for the reconstruction of pseudo-ZDR in small networks.
The reconstructed values are mostly too small. In addition, jumps appear in the time
series every time a satellite rises or sets. The magnitude of the jumps can be reduced
by downweighting of low elevation satellites; however, the reconstruction process
cannot be significantly improved therewith. In consequence, for analysis of satellite
or station specific effects in post-fit residuals we recommend undifferenced GNSS
data processing strategies, especially in small GNSS networks.
3.3.7
Tropospheric Parameters from Multi-GNSS Analysis7
P. Václavovic
Geodetic Observatory Pecný, Research Institute of Geodesy, Topography and
Cartography, Zdiby, Czech Republic
e-mail: [email protected]
J. Douša
Geodetic Observatory Pecný, RIGTC, Ondřejov, Czech Republic
e-mail: [email protected]
Nowadays, multi-GNSS offers new satellites and signals which are expected to
strengthen all estimated parameters, in particular the ZTD and horizontal linear
tropospheric gradients, or to densify slant tropospheric delays for monitoring the
troposphere asymmetry at individual GNSS sites. Currently, data from the US
NAVSTAR Global Positioning System (GPS) and the Russian GLONASS constellation are commonly used to produce different products within scientific services.
Essential models for these two systems has been already established, and their
mutual combination provides better precision then the processing from any
standalone system. Besides others, limitations for the use of GNSS data from other
global systems, the European Galileo and the Chinese BeiDou, persist mainly in
(1) incompleteness of the constellations, (2) lack of precise models and calibrations
for new signals, receiver and satellite instrumentations, and (3) lack of precise orbit
and clock products supporting the ultra-fast processing mode. The situation will
7
Parts from this section were previously published in Douša et al. (2018a).
3 Advanced GNSS Processing Techniques (Working Group 1)
87
change soon as both global systems will become operational in next years and the
IGS Multi-GNSS Experiment (MGEX, http://mgex.igs.org, Montenbruck et al.
2017) is continuously filling the gaps in data, metadata, models, formats, standards
and products for an optimal exploitation of all global satellite constellations and their
regional augmentations.
3.3.7.1
Evaluation of Results from Collocated GNSS Stations
The impact of using multi-constellation data on the tropospheric parameter estimation can be optimally assessed using closely collocated GNSS stations, e.g. within
few meters. Although different instrumentation-specific effects, such as phase centre
modelling and the quality of a receiver tracking, can affect analyses at both stations,
the station should principally observe the same tropospheric delays. For the purpose
of our evaluation, we selected two IGS station pairs, ZIM2-ZIMJ (Zimmerwald,
Switzerland) and MAT1-MATE (Matera, Italy), all collecting data from GPS,
GLONASS and Galileo systems within 10 m and 2.5 m in horizontal and vertical
distances, respectively. We used the GFZ MGEX (GBM) product (Deng et al. 2017)
and CODE MGEX (COM) product (Prange et al. 2017) as multi-GNSS reference
solutions for the comparison. We assessed not only the impact of using more
constellations but also the impact of different strategies for the parameter estimation.
The first approach is the Kalman filter usable mainly for real-time, and the second
strategy is the backward smoothing designed for improving the precision of parameters in post-processing (Václavovic and Douša 2015).
Solutions improved by introducing multi-constellations and the backward
smoothing are demonstrated in the time series of ZTD and horizontal gradient
differences obtained from the two collocated stations, ZIM2 and ZIMJ, Fig. 3.28.
Results of the single /multi-constellations are visualized by different colours:
(1) standalone GPS in red, (2) GPS + GLO in green, and (3) GPS + GLO + GAL
in blue. A positive effect is visible for all parameters, and is similar using both the
Kalman filter and the backward smoothing, i.e. for both real-time and postprocessing strategy. Scatters of multi-constellation solutions are smaller compared
to the standalone GPS solution. More significant effect is visible for the smoothed
gradient parameters. Theoretically, zero differences are expected for the collocated
stations with the same antenna height. However, a vertical difference between ZIM2
and ZIMJ is about 2 m which can cause about 0.5 mm difference in ZTDs when
considering the pressure decreases approximately by 11.3 Pa/m near the geoid and
the 100 Pa difference in the atmospheric pressure causes a 2.27 mm difference in
ZHD (Saastamoinen 1972). As we observe a ZTD difference about 2–3 mm, it can
be still attributed to remaining station-specific systematic errors, e.g. such as phase
centre offset and variation models.
Numerical statistics (biases and standard deviations) characterize an impact of
single- and multi-constellation solutions on the estimated ZTDs and horizontal linear
gradients when using the Kalman filter, Table 3.7. It should be noted, that
GLONASS (R) and Galileo (E) observations were down-weighted by a factor of
88
Fig. 3.28 Time series of
ZTD (top), north gradient
(middle) and east gradient
(bottom) differences at
Zimmerwald dual-station
when using the Kalman
filter (left) and the backward
smoothing (right)
J. Douša et al.
3 Advanced GNSS Processing Techniques (Working Group 1)
89
Table 3.7 Statistics (BIAS SDEV) for Kalman filter using GPS (G), GLONASS (R) and Galileo
(E)
Station
pair
ZIM2ZIMJ
ZIM2ZIMJ
ZIM2ZIMJ
MAT1MATE
MAT1MATE
MAT1MATE
GNSS
G
BIAS SDEV ZTD
[mm]
+2.8 1.4
BIAS SDEV N-GRD
[mm]
+0.08 0.17
GR
+2.4 1.3
+0.02 0.14
0.02 0.12
GRE
+2.0 1.3
+0.03 0.14
0.04 0.13
0.5 2.4
0.03 0.18
+0.18 0.25
GR
+0.1 2.3
+0.01 0.15
+0.14 0.22
GRE
+0.1 2.2
+0.00 0.15
+0.13 0.21
G
BIAS SDEV E-GRD
[mm]
0.02 0.14
2 with respect to GPS (G) to reflect lower accuracy of precise products and models.
A positive effect of multi-constellation is visible at all parameters, and particularly in
terms of the standard deviation, while the impact of GLONASS is more significant
compared to Galileo. It is expected due to a lower number of operational Galileo
satellites as well as longer support of GLONASS with precise models and products
in the scientific community. As already discussed, ZIM2-ZIMJ differences indicate a
bias of about 2–3 mm in ZTD which has been decreased partly in multi-GNSS
solutions. The improvements in all parameters reached 15–30% in terms of RMSE at
both dual-stations. Table 3.8 then shows the impact of the backward smoothing on
all solutions using single- or multi-constellation data. All the above mentioned
characteristics are similar to the Kalman filter, and the backward smoothing then
improved mainly standard deviations (by about 25%).
3.3.7.2
Carrier-Phase Post-fit Residuals and Slant Delays
Figure 3.29 shows the carrier-phase post-fit residuals when using the Kalman filter
PPP (left) and the backward smoothing PPP (right) for multi-GNSS solutions
supported with the COM (top) and GBM (bottom) MGEX products. The carrierphase residuals are useful indicators of an overall performance of the solution
including the quality of input products and models. Showing plots for the ZIM2
station only, below discussed characteristics are common to other stations too. First,
we observe a common elevation-dependent pattern of characteristics of post-fit
residuals when using elevation-dependent observation weighting. Second, the backward smoothing does not change the distribution of the carrier-phase post-fit residuals significantly. The main effect of the backward smoothing is thus understood
mainly as improved accuracy of the estimated parameters. The tropospheric slant
delays reconstructed from the model parameters and post-fit residuals will thus
90
J. Douša et al.
Table 3.8 Statistics (BIAS SDEV) for backward smoothing using GPS (G), GLONASS (R),
Galileo (E)
Station
pair
ZIM2ZIMJ
ZIM2ZIMJ
ZIM2ZIMJ
MAT1MATE
MAT1MATE
MAT1MATE
GNSS
G
BIAS SDEV ZTD
[mm]
+2.7 1.1
BIAS SDEV N-GRD
[mm]
+0.11 0.12
GR
+2.3 1.0
+0.06 0.11
0.02 0.09
GRE
+1.9 1.0
+0.07 0.12
0.04 0.09
1.3 1.6
0.04 0.15
+0.22 0.19
GR
+0.6 1.4
+0.00 0.12
+0.16 0.17
GRE
+0.5 1.4
0.01 0.11
+0.16 0.16
G
BIAS SDEV E-GRD
[mm]
0.02 0.10
Fig. 3.29 Carrier-phase post-fit residuals from the Kalman filter (left) and the backward smoothing
(right)
benefit primarily from the improvement of the parameters. Third, the GPS residuals
(black) are the smallest and compact compared to other systems indicating actual
quality of precise models and products. Galileo shows the largest residuals, however,
we had to substitute various precise models, in particular station antenna phase
centre offsets and variations by using the values from GPS models. Due to the
same reason, we may notice systematic changes in the elevation-dependent redistribution of Galileo residuals (red) after applying the backward smoothing. Fourth, we
can notice about twice larger post-fit residuals GLONASS (green) when using the
GBM product compared to the COM product. As the characteristics are common to
all the stations, it indicates a lower quality of GLONASS orbits and clocks from the
3 Advanced GNSS Processing Techniques (Working Group 1)
91
GBM product or some inconsistent models used for the product generation and in the
PPP software.
3.3.8
Multi-GNSS Solutions and Products
Z. Deng
GFZ German Research Centre for Geosciences, Potsdam, Germany
e-mail: [email protected]
GPS PWV is considered to have observation noise of about 1~2 mm, but GPS PWV
sometimes shows larger noise and jumps for some stations and/or at certain times,
suggesting that a lower number of GPS satellites and poor line-of-sight condition
limits the quality of ZTD estimates in such stations and at such times. Many GPS
networks are now being upgraded to multi-GNSS observation networks, and this
upgrade is expected to be beneficial for GNSS tropospheric monitoring. The IGS
network is being upgraded to be capable to observe multi-GNSS (GPS, GLONASS,
Galileo, BeiDou and QZSS) signals. GFZ has started to provide multi-GNSS orbit
and clock for all the constellations (GBM) since middle 2014 (Fig. 3.30).
To validate ZTD results from a global GPS-only and multi-GNSS analysis we
processed multi-GNSS station data from the global ground tracking MGEX-network
(Fig. 3.31) spanning a 3-month time period from December 2014 to February 2015.
Color indicates maximum number of satellites available in addition to GPS per
observation epoch.
Multi-GNSS and GPS-only global network solutions were generated to study the
impact of including additional GNSS on estimated ZTD. ZTD difference time series
Fig. 3.30 Number of satellites per GNSS constellation included in the global multi-GNSS observation data processing
92
J. Douša et al.
Fig. 3.31 Site selection for multi-GNSS processing derived from pre-defined sites associated to
GFZ Rapid routine solution
were computed for sites which provided multi-GNSS observation data. Dots depict
sites suitable for comparing GPS-only versus multi-GNSS (GREC) results. Figure 3.32 shows the mean bias of the resulting ZTD differences which varies in the
range of 1.5 mm. No specific latitude or longitude dependency could be identified.
Moreover, there is no obvious correlation between the magnitude of the bias and the
number of additional satellites used to estimate the ZTDs (compare to Fig. 3.31).
Figure 3.33 shows associated standard deviations derived from the ZTD difference
time series with values below 3.5 mm (Deng et al. 2015).
ZTDs estimated with multi-GNSS processing are more stable than those based on
GPS only. Sudden jumps observed in GPS-only ZTD are significantly reduced with
the multi-GNSS processing. Because the number of satellites in multi-GNSS solution is more than twice that of only GPS observation, the noise due to rising and
setting satellites is mitigated thus reducing the size of sudden jumps in ZTD.
3.4
PPP and Ultra-Fast GNSS Tropospheric Products
A majority of E-GVAP ACs till now uses a double-difference observation
processing in the network solution. This strategy eliminates clock errors at GNSS
receiver and satellite while public products were not available in near real-time
(NRT). The situation has changed in 2012, when the IGS introduced the Real-Time
Service (RTS, http://rts.igs.org) providing GPS and GLONASS orbit and clock
corrections by combining contributions from several IGS real-time analysis centres
3 Advanced GNSS Processing Techniques (Working Group 1)
93
Fig. 3.32 Mean bias for zenith total delay estimates between GPS-only and multi-GNSS (GRCE)
solution. Differences are shown only for those sites actually providing multi-GNSS observation.
Left and top subfigures show mean bias distribution w.r.t. latitude and longitude, respectively
Fig. 3.33 Shows associated standard deviations (StDev) derived from the ZTD difference time
series with values below 3.5 mm. In contradiction to the biases, the STDs reveal a small latitude
dependency with larger magnitudes for sites below the equator
94
J. Douša et al.
(Caissy et al. 2012). The IGS RTS aims at supporting real-time (RT) analyses with
the Precise Point Positioning (PPP) method (Zumberge et al. 1997). The PPP is
based on original observations or their linear combination without differencing
between receivers or satellites. Though German Research Centre for Geosciences
(GFZ) has provided a NRT PPP ZTD product (Dick et al. 2001; Gendt et al. 2004)
since 2001, it was possible only thanks to their two-step processing approach
consisting of (1) a global NRT solution for determining consistent satellite clock
and orbit products and (2) a distributed PPP processing for ZTD estimated for each
station individually. With the availability of global real-time data flow, software and
standards specified for precise product dissemination, the PPP is becoming more
popular for the troposphere monitoring.
Compared to the traditional approach in E-GVAP dominated by the doubledifference network processing, the PPP offers several advantages: (a) an easy
production in real-time or NRT fashion, (b) flexible use of central or distributed
processing scheme including a receiver built-in solution, (c) an estimation of tropospheric parameters in the absolute sense with a high spatio-temporal resolution, and
(d) an optimal support of all satellite constellations and new signals including
multiple frequencies; all profiting from a highly efficient and autonomous processing
approach. The price for mentioned advantages is however paid by several disadvantages. Compared to the strategy using double differences, all observation models
need to be carefully applied to reach the best accuracy. In addition, integer ambiguity
resolution is possible only if precise observation phase biases are available, thus
often non-integer-fixed ambiguities are usually estimated.
3.4.1
Real-Time Data and Product Dissemination
Y. Altiner
BKG, Federal Agency for Cartography and Geodesy, Frankfurt, Germany
e-mail: [email protected]
W. Söhne
BKG, Federal Agency for Cartography and Geodesy, Frankfurt, Germany
e-mail: [email protected]
A. Stürze
BKG, Federal Agency for Cartography and Geodesy, Frankfurt, Germany
e-mail: [email protected]
PPP is a method for high accuracy positioning using observations from a single
GNSS receiver, suited for both, real-time and post-processing implementations.
Traditionally, PPP is an idea of a post-processing technique for efficient evaluation
of GNSS data from large scale networks. But it also enables a real-time positioning
for stable (static) and movable (kinematic) objects with an accuracy of centimeter
and sub-decimeter level, respectively. The fundamental advantage of PPP is that the
3 Advanced GNSS Processing Techniques (Working Group 1)
95
number of simultaneously observed stations within a global network can be significantly increased without decrease in accuracy of station coordinates.
This benefit and the short duration of data processing has given the PPP method
popularity to be used it efficiently also in the field of climate research, in particular to
support weather forecasting techniques for medium or large scale areas. Using PPP,
estimation of coordinates for positioning and determination of ZTD is possible every
second (real-time PPP). However, for PPP in real-time some parameters, such as
precise satellite coordinates (orbit), earth orientation parameters, and satellite clock
corrections are needed from an external source, e.g., available online from the
IGS RTS.
The IGS RTS is generating and providing the variables of the SSR related to the
orbits and clocks of GNSS satellites, an indispensable essential for real-time PPP
(http://www.igs.org/rts). The RTS products created by several analysis centres
contain GNSS satellite orbit and clock corrections to the broadcast ephemeris.
Orbit corrections are provided as along-track, cross-track and radial offsets to the
Broadcast Ephemeris in the ECF reference frame (Earth-centred and Earth-fixed).
RTS corrected orbits are expressed within the ITRF implemented during the realtime GNSS observations. Clock corrections are expressed as offsets to the Broadcast
Ephemeris satellite. Hereby, attention should be paid that the reference point of the
satellite clocks is selected in accordance with the reference point of the satellite
orbits.
The IGS RTS is providing three combination solutions, IGS01, IGS02 and
IGS03. While IGS01 is generated on the basis of epoch-wise combination, the
Kalman filter technique is exploited for producing IGS02 and IGS03. Two different
agencies are responsible for these RT products: European Space Agency (ESA)
provides IGS01 and BKG provides IGS02 and IGS03. All mentioned streams
include orbits/clock corrections to GPS satellites, and only IGS03 supports also
GLONASS constellation. While IGS01 and IGS02 are combined from up to eight
individual solutions, IGS03 has only four individual contributors.
The RTS correction streams are formatted with respect to the RTCM (Radio
Technical Commission for Maritime Services) standard for SSR and are transmitted
using the NTRIP protocol (Networked Transport of RTCM via Internet Protocol).
NTRIP was developed in co-operation between the Informatikzentrum Dortmund in
Germany and BKG (http://igs.bkg.bund.de/ntrip) and initiated as an industrial standard since 2004 (Weber et al. 2005). Afterwards, NTRIP was standardized by the
Special Committee 104 “DGNSS” of RTCM. The communication between the
major components of the NTRIP, i.e. the server, the caster and the client are handled
through HTTP ports. It is to mention that the major software components of the
NTRIP are developed under “GNU General Public License”. NTRIP allows disseminating hundreds of data and product streams simultaneously for a few thousand
users when applying the modified Internet Radio Broadcasting Software. It is also to
note that a GNSS stream typically needs not more than 5 kbit/s bandwidth. The
currently used version 2 of NTRIP, downward compatible to version 1, was completed in 2009. BKG supports the distribution of the new technology by providing
the so-called Professional NTRIP Caster. This tool has been developed in
96
J. Douša et al.
cooperation of BKG with Alberding Company in Germany for administration,
configuration, and implementation the piece of software running using the LINUX
operating system and is widely used within the IGS and EUREF (http://www.epncb.
oma.be/). Meanwhile, almost every new GNSS receiver is coming with the NTRIP
option.
3.4.2
BKG Real-Time Analysis Development
and Contribution
Y. Altiner
BKG, Federal Agency for Cartography and Geodesy, Frankfurt, Germany
e-mail: [email protected]
W. Söhne
BKG, Federal Agency for Cartography and Geodesy, Frankfurt, Germany
e-mail: [email protected]
A. Stürze
BKG, Federal Agency for Cartography and Geodesy, Frankfurt, Germany
e-mail: [email protected]
Since April 2016, BKG is contributing to the COST real-time demonstration campaign for troposphere estimation providing solutions processed with BNC, BKG’s
own software tool. The development of BNC started in 2005 by BKG in collaboration with different partners, e.g. with Technical University Prague, Czech Republic,
and Alberding Company, Germany. BNC is a software for simultaneously retrieving, decoding, converting and processing or analyzing real-time GNSS data streams
applying the NTRIP standard. BNC has been developed within the framework of
EUREF and the IGS. Although BNC is primarily intended for real-time GNSS
applications, it may also be run offline by transmission of data from an external
file to simulate real-time observation conditions or for post-processing
implementations. BNC provides also different modes of data evaluation like
“graphics or interactive” mode to illustrate the processing state and results and “no
windows” mode as well.
A major module of the BNC is the option “Real-Time PPP” for positioning in
real-time according to the SSR model which was first provided in 2010. To meet
requirements of the PPP in real-time using the state variables of the SSR model,
within this version additional messages are provided to the user, among others,
satellite orbit and clock corrections for GPS as well as GPS and GLONASS
combination, and ionospheric corrections as well as biases for code and phase
data. The PPP module of BKG allows users a real-time positioning worldwide at
sub-decimeter-level using code and phase data in ionosphere-free solutions using P3
or L3 linear combinations in static or kinematic mode within an observation time of
10 min. In 2011, the “Real-Time PPP” module was expanded by incorporating the
3 Advanced GNSS Processing Techniques (Working Group 1)
97
PPP implementation for post-processing to work offline including data from external
files. In 2014, BNC comes into being able to process data of multiple stations in a
single BNC job (one job for all stations). Within this version (BNC 2.12), the
troposphere parameters estimated by the data processing is provided by SINEXTRO format (v0.01) to allow the usage of the TRO-file as a priori or a posteriori
model for GNSS applications (Weber et al. 2016).
Within the RT demonstration campaign, ZTD values were evaluated in 5-min
intervals for 22 mount-points at the beginning. As of October 2017, all 32 mountpoints are implemented. As agreed, orbit and clock corrections from the IGS03
product are used. Streaming of broadcast ephemeris is taking place through the realtime access to the broadcast ephemeris stream of BKG “RTCM3EPH” for GPS and
GLONASS. It should be noted that the IGS03 correction stream is a combined
stream of four individual solutions providing GPS and GLONASS and is done by
BKG with BNC.
The evaluation of ZTD implemented by BNC is using the Kalman filtering
method. One important parameter to set in the configuration file is the white noise
(signal) for 1 h which can be according to the weather conditions. The standard value
for the variation of the white noise is 36 mm/h1/2. Other effects or parameters
influencing the accuracy of positioning so far considered by BNC are listed in
Table 3.9. Processing of data within the demonstration campaign takes place using
an elevation cut-off angle of 7 for observation usage and considering the float
ambiguity resolution.
To study the impact assessment of GNSS processing on tropospheric products,
two different products are created by BKG. A GPS-only and a combined GPS plus
GLONASS product is computed and submitted separately as GPS and GPS + GLO
products on hourly basis to the central analysis centre of the project (http://www.
pecny.cz/COST/RT-TROPO/). Each hourly solution contains 12 ZTD-values and
each ZTD-value represents a time span of 5 min within the relevant hourly solution.
In total, 10,234 files were submitted on hourly basis from March 14, 2016 to July
31, 2017 including solutions using the real-time GPS measurements and state
Table 3.9 Important effects so far considered by BNC within data processing
Effect
Earth’s tide
Earth’s rotation (Movement of pol)
Phase-wind up
Ionosphere
Troposphere
Multipath (phase shift of the signal)
Atmospheric and hydrological loading
Ocean tides
Offset and phase centre variations of satellite and ground
antennas (PCV)
Considered
Yes
Yes
Yes
First order terms eliminated
using L3
Determined
No
No
No
Yes
98
J. Douša et al.
variables from the IGS03 product stream. For GPS + GLO, the quantity of the
submitted hourly files reached to a total number of 8988 between March 18, 2016
and July 31, 2017. Data transmission continues also beyond the date 31.7.2017
(Fig. 3.34).
The internal precision in terms of agreement of both BKG solutions – GPS-only
versus GPS + GLONASS – is shown in Figs. 3.35, 3.36 and 3.37. The example
covering a 15-day period in July 2017 shows a good overall agreement for the ZTD
Fig. 3.34 The quantity of ZTD processed by PPP using orbit and clock corrections from IGS03
product for GPS and GPS + GLO observations. The solutions were created in 5-min intervals and
combined to a single file on hourly basis to be submitted to the central analysis centre of the project
(http://www.pecny.cz/COST/RT-TROPO/). Each submitted hourly solution contains
12 ZTD-values and each ZTD-value represents a time span of 5 min within the relevant hourly
solution
Fig. 3.35 Time series of ZTD estimates from BKG’s real-time solutions (GPS-only and
GPS + GLONASS) as taken from the uploaded COST format files for stations THTI (left) and
WTZR (right)
3 Advanced GNSS Processing Techniques (Working Group 1)
99
Fig. 3.36 Time series of differences of ZTD estimates between BKG’s real-time solutions “GPSonly” minus “GPS + GLONASS” for stations THTI (left) and WTZR (right). The plots are showing
increased differences during periods of re-initialization of at least one of both solutions
Fig. 3.37 Mean of differences between time series of ZTD estimates from BKG’s real-time
solutions GPS-only minus GPS + GLONASS for 15 days in July 2017. Only three of 21 stations
show a bias of almost 2 mm ZTD. The large standard deviation for station NRMD cannot be
explained by frequent re-initialization; the time series show larger portions of disagreement
time series for Tahiti (Fig. 3.35 left) and Wettzell. A closer look into the differences,
however, shows some portion of disagreement. This can be explained by
re-initialization of at least one of both solutions. Figure 3.37 shows the statistics
for 21 stations (station ADIS was only occasionally available in the GPS + GLO
solution). For the majority of the stations the mean bias is well below 2 mm ZTD.
Regarding the submitted ZTDs of BKG within the demonstration campaign, a
statistical study for station WTZR was conducted in relation to the EUREF weekly
100
J. Douša et al.
combined solutions (Table 3.10). To do this, the EUREF weekly combined ZTD
solutions were considered as target values and BKG real-time GPS and GPS + GLO
solutions as measured values, respectively. The measured values from BKG were
subtracted from the EUREF target values. In total, 8451 ZTD values from each
solution were included to this statistical study. The correlation between the differences of GPS and GNSS solutions in relation to the EUREF combined solutions
amounts to 85%. This suggests good coincidence between both solutions (GPS and
GPS + GLO). The average differences for GPS and GPS + GLO are on the order of
11.4 mm and 8.6 mm, respectively. Contrary to the average, the standard deviation
of GPS is smaller than the standard deviation of GPS + GLO (11.4 mm and 12.9 mm,
respectively).
To illustrate the relation of the GPS and GPS + GLO solutions to the EUREF
combined solution the total 8451 ZTD values were reduced to 253 through choosing
a random value for each day at 15:30 as illustrated in Fig. 3.38. It is important here to
Table 3.10 Statistical aspects between ZTDs determined by BKG for station WTZR using GPS
and GPS + GLO real-time corrections and the combined EUREF weekly solution. In total, 8451
ZTD values from each solution were included in the study. Averages and standard deviations,
shown in bold, determined relative to the EUREF combined ZTD solution (EUREF minus
BKG-GPS and EUREF minus BKG-GNSS)
Solution for station WTZR
BKG-GPS
BKG-GPS + GLO
EUREF
Average in mm
11,8
8,1
Std. dev. in mm
11,4
12,9
Average of std. dev. in mm
0,7
0,7
0,5
Fig. 3.38 Relation of the GPS and GPS + GLO solutions to the EUREF combined solution for
station WTZR. 8451 ZTD values from each solution were reduced to 253 through choosing of a
random value for each day at 15:30. The results suggest good coincidence between the GPS and
GPS + GLO solutions using the real-time orbit and clock corrections
3 Advanced GNSS Processing Techniques (Working Group 1)
101
notice that the EUREF weekly combinations include the results of several ACs
which are derived using the final orbits of the IGS. Each AC does not use the same
software for data processing, and network geometries of ACs are also different.
Some studies conducted by BKG between the results of the real-time, near real-time
und post-processing applications also suggest the good quality of real-time PPP with
respect to the determination of ZTD (Altiner et al. 2009, 2010 and 2011).
3.4.3
Assessment of IGS RTS Orbits and Clock Corrections
and GOP Real-Time Tropospheric Products8
L. Zhao
Geodetic Observatory Pecný, Research Institute of Geodesy, Topography and
Cartography, Zdiby, Czech Republic
e-mail: [email protected]
P. Václavovic
Geodetic Observatory Pecný, Research Institute of Geodesy, Topography and
Cartography, Zdiby, Czech Republic
e-mail: [email protected]
J. Douša
Geodetic Observatory Pecný, RIGTC, Ondřejov, Czech Republic
e-mail: [email protected]
The accuracy of the RT ZTD calculated with the PPP method strongly depends on
the quality of RT GNSS orbit and clock corrections (Douša and Václavovic 2014;
Hadaś and Bosy 2015). We evaluated publicly available global RT products and we
summarized our ZTD contributions to the RT Demonstration Campaign initiated in
2015 by this COST Action. We also studied the impact of IGS RTS (Caissy et al.
2012) on the simulated RT ZTD estimates within the GNSS4SWEC Benchmark
Campaign (Douša et al. 2016).
3.4.3.1
Assessment of Real-Time Orbit and Clock Corrections
We investigated the performance of four real-time products (Douša et al. 2018a)
having been collected and archived at Geodetic Observatory Pecny (GOP) using the
BNC Software (Weber et al. 2016) since 2013: IGS01, IGS02 and IGS03 the official
IGS RTS combined products, and CNS91 (also known as CLK91) as an individual
solution provided by CNES RT AC (Laurichesse 2011). Two different strategies and
8
Parts from this section were previously published in Douša et al. (2018a).
102
J. Douša et al.
Fig. 3.39 Monthly statistics of availability of satellite corrections from IGS01 RT stream
software are used for combining the IGS RTS (Hadaś and Bosy 2015). All mentioned streams include orbits/clock corrections to GPS satellites, and only IGS03
supports also GLONASS constellation. Navigation data from MGEX (Montenbruck
et al. 2017) needs to be used together with the RT corrections to recover the precise
satellite orbit and clocks. First, the availability and completeness of RT corrections
were checked and, second, satellite orbit and clocks were compared to IGS final orbit
and clocks, both during the period of 2013–2017.
Figure 3.39 depicts monthly completeness of RT corrections for all GPS satellites
from the IGS01 combined products. The others IGS products and the CNES product
are generally showing similar performance. From the comparison, we can classify
problems into three groups: (1) temporal unavailability period of some satellites,
e.g. G03, G04, (2) source-specific unavailability, e.g. G01 for CNS91, and (3) satellite-specific incompleteness. The first group is usually caused by the loss of observations due to the upgrade of satellite, such as replacing the old Block IIA satellites
with the new Block IIF satellites or a maintenance identified by satellite unhealthy
status. The second and third groups of gaps are caused by data unavailability from a
global network and the processing strategy including outlier detection in the product
generation. The availability of the corrections is significantly lower for some months
(June 2015, December 2016) compared to others which was caused by the internet
connection failures at GOP when receiving the streams. The source-specific loss of
data at IGS02 and CNS91 streams are visible in June 2015 and, these are mainly due
to the inconsistent navigation message Issue of Date (IOD) available from the
MGEX broadcast and those referred by RT corrections. It can be thus recommended
to use consistent RT navigation data and precise correction streams optimally
guaranteed by the same provider. In general, the availability of RT corrections is
well over 90% for most satellites which agrees with findings in (Hadaś and Bosy
2015). It indicates that the RT corrections were provided continuously for use in
troposphere monitoring, however, problems can be expected in a kinematic positioning which is more sensitive to the product incompleteness.
3 Advanced GNSS Processing Techniques (Working Group 1)
103
Table 3.11 RT clocks and orbits components compared to the IGS final products
IGS01
IGS02
IGS03
CNS91
RMSE [mm]
Radial
1.84
2.35
2.41
2.68
Along
2.83
3.71
3.82
3.07
Cross
2.38
3.04
3.10
2.47
3D
4.34
5.63
5.70
5.01
Clock
5.72
10.08
10.28
11.16
SDEV [mm]
Clock
2.95
3.52
3.03
2.29
Apart from the availability of the corrections, the precision is critical for the user
performance. The orbits are compared in 5-min intervals for three components:
radial, along-track and cross-track while the clock comparison is based on the second
order difference method. The IGS08 and the IGS14 model is used to correct satellite
PCOs prior and after January 29, 2017, respectively, corresponding to the adoption
of the IGS14 reference frame (Rebischung et al. 2016). The clock datum is estimated
by calculating average over all satellites clocks at each epoch. The datum inconsistencies are then eliminated through single-differences between individual satellite
clocks and the clock datum. The single-differences from the real-time clocks are
compared to those from the IGS final product. The root-mean-square error (RMSE)
of RT orbits and clocks are calculated for each day while outliers are removed using
a fixed threshold. Although there is a strong correlation between clocks and radial
orbit component, we haven’t corrected this dependency. Table 3.11 gives summary
statistics for all products over all days.
The orbit difference in radial component shows the smallest RMSE for all
products, whereas the along-track and cross-track components reached slightly
larger values. The IGS01 orbit shows the best agreement with respect to the IGS
final orbits. Largest differences are observed for the orbits from IGS03, which might
be attributed to a different outlier detection method applied when including
GLONASS satellites. Time evolution of the orbit comparison for each product and
specific component is shown in Fig. 3.40. Coordinate differences greater than 30 cm
are plotted at the top horizontal lines of each graph. Orbits from the IGS01 stream are
less affected by the outliers compared to IGS02 and IGS03 products as indicated by
outliers mainly during March 2015. The switch from the IGS08 to the IGS14 PCO
model (January 28, 2017) can be observed in statistics of the radial component. It
seems that the CNS91 product used the new IGS14 model as of March 9, 2017,
while official IGS solutions are difficult to recognize due to most likely asynchronous switches by different contributing providers. Otherwise, the orbit accuracy for
all products shows an overall good consistency over the period.
Table 3.11 also summarizes RMSE and standard deviation (SDEV) of the realtime clock corrections. The former represents the accuracy relevant for the
processing of code pseudoranges while the latter characterizes the precision important for the carrier-phase processing. It can be also interpreted from the PPP point of
view combining both observation types as follows – the former have a positive
impact on the PPP convergence time while the latter enable more precise positioning
within already converged solution (Ye et al. 2018). Obviously, this is the case of
104
J. Douša et al.
Fig. 3.40 Daily RMS of real-time orbits with respect to IGS final orbit
Fig. 3.41 Daily statistics SDEV (top) and RMSE (bottom) of real-time clocks
IGS01 and CNS91 products when the first is more accurate, but the second more
precise for the PPP application. The IGS02 and IGS03 products performs slightly
worse in terms of both RMSE and SDEV.
Figure 3.41 finally shows time series of the clock SDEV and RMSE statistics.
The former (top plot) indicates a comparable high quality over the period for IGS01
3 Advanced GNSS Processing Techniques (Working Group 1)
105
and CNS91, while more outliers are observed for IGS02 and IGS03 including the
problematic period in 2015 identified in the orbit availability evaluation. The clock
RMSE from IGS01 is the lowest and the most stable compared to the others during
the period while, the RMSE of CNS91 clocks was more accurate during 2015 when
compared to the other years.
3.4.3.2
Impact of IGS RTS Products on ZTD Estimates
The impact of the IGS RTS products on PPP ZTD estimates was assessed by
exploiting the GNSS4SWEC Benchmark campaign with 400 GNSS stations in
central Europe during the period of May–June, 2013. The ZTD was calculated
using the G-Nut/Tefnut software in the post-processing mode when supported
with two precise products: (1) IGS final orbits and clocks, and (2) IGS RTS orbit
and clock corrections. Two reference solutions were provided for the benchmark
using different software and processing strategies (Douša et al. 2016). GOP used the
BSW and the double-difference processing (DD) and GFZ used the EPOS software
and the PPP method. Statistics from the comparison of both testing solutions with
respect to both reference products are given in Table 3.12. Generally, the results
indicate a good agreement, however, the impact of the IGS RTS products (IGS01) on
ZTDs is clearly visible in two aspects: a) a common systematic error of 2.4–2.8 mm,
and b) a lower precision of 13–17%. Interestingly, a better agreement in terms of
SDEV is reached between 10% and 20% when using two PPP solutions (G-Nut/
Tefnut vs EPOS software) compared to the processing strategies (DD vs PPP). The
results also showed that input products and the processing strategy might result in a
similar impact on the ZTD estimates which can reach up to 20% in terms of
accuracy. Finally, it should be noted that the PPP ZTD estimation used a stochastic
model and an epoch-wise filtering method in the G-Nut/Tefnut software (Václavovic
et al. 2013), while a deterministic model with the least-squares batch adjustment
used in the EPOS software.
Table 3.12 Summary statistics from the comparison of PPP ZTD results using two inputs (IGS01
RT vs. IGS final) w.r.t. EUREF reprocessing
G-Nut/Tefnut PPP Input precise
products
IGS final (SP3 files)
IGS01 RT corrections
IGS final (SP3 files)
IGS01 RT corrections
ZTD reference
product
GOP final (BSW52/
DD)
GOP final (BSW52/
DD)
GFZ final (EPOS/
PPP)
GFZ final (EPOS/
PPP)
Bias
[mm]
+0.9
STD
[mm]
5.1
RMS
[mm]
5.2
+2.4
5.8
6.4
+0.4
4.1
4.2
+2.8
4.9
5.7
106
3.4.3.3
J. Douša et al.
Long-Term Quality of Operational RT ZTD Production
The RT ZTD from the demonstration campaign is evaluated for 18 European stations
during the initial year of the GNSS4SWEC Real-time Demonstration campaign.
Two GOP solutions using the IGS03 product are compared with respect to the
EUREF 2nd reprocessing combined tropospheric product (Pacione et al. 2017):
(1) GOPR – standalone GPS solution, and (2) GOPQ – GPS + GLONASS solution.
In Table 3.13, we can observe a systematic error in ZTD of about 2 mm in the longterm evaluation, similar as observed in the simulated real-time processing in the
benchmark campaign, see the previous subsections. Although GLONASS observations are down-weighted by a factor of 2 in our solution in order to reflect the lower
quality of GLONASS precise products, a small positive impact on the ZTD is
observed in terms of mean bias (10%) and mean SDEV (7%), both calculated over
18 stations.
Figure 3.42 shows the comparison of the GOPR solution with respect to the
EUREF combined product during the first year of the RT demonstration campaign.
Monthly mean ZTD biases, standard deviations and their 1-sigma scatter calculated
over all 18 stations indicate a long-term stability of the operational real-time
production with a small seasonal effect in SDEV due to a less accurate troposphere
modelling during the summer period (Douša and Václavovic 2016).
3.4.4
Real-Time Product Development and Evaluation
at ROB
E. Pottiaux
Royal Observatory of Belgium, Brussels, Belgium
e-mail: [email protected]
In the framework of COST Action ES1206 (GNSS4SWEC), the Royal Observatory
of Belgium (ROB) collaborated with the Geodetic Observatory Pecny (GOP) to use
their Real-Time Precise-Point-Positioning (RT-PPP) software G-Nut/Tefnut for the
real-time monitoring of the troposphere, to help support nowcasting of severe
weather in Belgium. On June 21, 2014, ROB started to use G-Nut/Tefnut to produce
real-time tropospheric products (ZTD and horizontal gradients) using 4 different
processing strategies (Table 3.14), with a particular focus on Belgium.
Table 3.13 Summary statistics over 18 stations from routine RT product using GPS and
GPS + GLO data
Solution
description
GOPQ –
GPS + GLO
GOPR – GPS
BIAS
[mm] mean sdev
1.8 2.9
SDEV
[mm] mean sdev
6.7 1.2
RMSE
[mm] mean sdev
7.5 2.5
2.0 2.8
7.2 1.0
7.9 1.5
3 Advanced GNSS Processing Techniques (Working Group 1)
107
Fig. 3.42 Monthly summary biases and standard deviations of real-time ZTDs over 18 stations
Table 3.14 Setup of the different ROB’s RT-PPP Processing
Common parameters to all solutions
Parameter
Setup
Coordinates
Static estimation
Tropospheric parameters
ZTD + horizontal gradients
Tropospheric model
Saastamoinen + GMF + Chen and Herring
Cut-off angle
3
Time resolution
10 s
Latency
100 s
Ocean tide loading Coef.
FES2004
Antenna model
IGS08 Antex file
Solution naming & differences
Solution name
GNSS observations
CLK + ORB product
ROBA
GPS-only
IGS02 (GPS only)
ROBB
GPS
IGS03 (GPS + GLONASS)
ROBC
GPS + GLONASS
IGS03 (GPS + GLONASS)
ROBD
GPS + GLONASS
CNS91 (GPS + GLONASS)
Based on these developments, ROB also participated in the RT-PPP demonstration campaign, with the main goals to extensively develop, test, and validate realtime processing methods and tropospheric products that can help supporting
nowcasting and forecasting of severe weather and foster the link to WG2 activities.
ROB processes thus real-time GNSS observations from the 32 GNSS sites requested
to participate in the demonstration campaign, along with those from 153 additional
GNSS sites located worldwide (Fig. 3.43, left), including the complete Belgian
dense network (Fig. 3.43, right). In total, 185 GNSS Stations are included in
ROB’s RT-PPP processing with G-Nut/Tefnut (Fig. 3.44). These 185 stations are
equipped with 73 different combinations of GNSS receivers and antennas
(26 receiver types, 40 antenna type, Fig. 3.45), allowing thereby to study and assess
the performances of this RT-PPP processing w.r.t. the equipment, and in fine to finetune accordingly the processing strategy. All RT-PPP products are formatted in both
108
J. Douša et al.
Fig. 3.43 (Left) Location of the GNSS stations included in the RT-PPP operated by ROB using the
G-Nut/Tefnut software from GOP. (Right): The location of the GNSS stations of the Belgian Dense
network included in this processing
Fig. 3.44 Number of GNSS stations included in the RT-PPP Processing operated by ROB
Fig. 3.45 Number of stations equipped with a specific type of GNSS receiver (left) or antenna
(right) and included in ROB’s operational RT-PPP processing campaign
3 Advanced GNSS Processing Techniques (Working Group 1)
109
in the COST-716 v2.2a and in the SINEX_TRO v2.00 format to ease exchange and
validation. Upload of these products (only the stations contributing to the WG1 RT
demonstration campaign) is done every hour to a central hub at GOP, and can be
visualized at http://www.pecny.cz/COST/RT-TROPO.
3.4.4.1
Monitoring and Validation
Developing and maintaining such new RT-PPP processing systems and products
requires regular assessment, adaptation, and fine-tuning cycles. Therefore, ROB
developed its own monitoring/validation system consisting of MySQL databases,
statistical assessment programs, and a web-based user-interface to monitor continuously his RT-PPP processing system (graphs, reports. . .). The system is capable of
monitoring/carrying out:
• The campaign setup and its evolution
• Inconsistency checks (configuration, equipment, models. . .)
• The performances of the products at all GNSS stations included in the RT-PPP
processing (the monitoring system developed by GOP can only monitor 17% of
them), at various time scale and epochs (biases, precision, geographical
dependency. . .)
• Alarm systems in case of problem (dataflow, processing. . .)
To validate the performances of the products at all stations and all time scales
(e.g. a very rapid monitoring require the use of e.g. NRT products), the monitoring
system uses various reference products listed in Table 3.15, from post-processed to
NRT, some computed in PPP, some in Double-Difference (DD) approach etc. In all
cases, the validation starts with a screening of the RT-PPP results to reject convergence period. This screening is using the RMS of the coordinates, ZTD and
horizontal gradients, as well as the GDOP values, and the number of satellite
measurements used to compute the tropospheric products at each single epoch.
The advantages of this approach is that it can be implemented as a “real-time
Table 3.15 List of reference products chosen to validate the RT-PPP products operationally
computed by ROB
Solution
IGS
Software
BSW50
ROB PPP
BSW52
ROB DD
BSW52
ROB NRT
E-GVAP
BSW52
Orb. &
Clk.
IGS
final
CODE
final
CODE
final
IGS
ultrarapid
MultiGNSS
GPS only
Time
Res.
5 min
Latency
3–4 weeks
NB.
Sta.
30
TRO Est.
ZTD + GRD
GPS + GLO
5 min
3 weeks
All
ZTD + GRD
GPS + GLO
1h
3 weeks
All
ZTD + GRD
GPS only
15 min
1h
All
ZTD only
110
J. Douša et al.
filtering” at the production level to screen out potential performance degradations
before using the product in actual meteorological applications. In most cases, this
filtering works well and had a low percentage of rejected values (<1%).
Having various reference products is justified by their respective advantages and
drawbacks, namely IGS for a standard reference product (but having only 30 common stations), consistency for PPP approaches (but probably less precise/accurate
than the double-difference approach), accuracy/precision for final products (but
large latency of the monitoring), and short-latency monitoring in the case of NRT
solutions.
The first long-term validation has been carried out with this system over the
period July 1st 2014–March 31st 2015 (9 months). Globally and at the long-term
level, the 4 solutions listed in Table 3.14 agrees very well with the IGS Final
products (22 stations considered altogether). The linear regressions (computed
over ~600,000 samples) between RT-PPP and the IGS products have a correlation
coefficient above 0.99, a slope of ~0.99, and almost a zero intercept (0.02–0.03 mm).
The bias is however station dependent, consistent at the mm level for all products,
and ranges from 5 to 16 mm (Fig. 3.46). The standard deviation is also station
dependent and ranges typically from 5 to 10 mm (Fig. 3.47), with a maximal value
Fig. 3.46 Bias observed between the ZTDs from each ROB’s RT-PPP product and the IGS final
troposphere product over the period July 1st 2014–March 31st 2015
Fig. 3.47 Standard deviation observed between the ZTDs from each ROB’s RT-PPP product and
the IGS final troposphere product over the period July 1st 2014–March 31st 2015
3 Advanced GNSS Processing Techniques (Working Group 1)
111
Fig. 3.48 Bias observed between the ZTDs from each ROB’s RT-PPP product and the ROB final
PPP troposphere product over the period July 1st 2014–March 31st 2015. Stations are ordered by
increasing longitude
Fig. 3.49 Standard deviation observed between the ZTDs from each ROB’s RT-PPP product and
the ROB final PPP troposphere product over the period July 1st 2014–March 31st 2015. Stations are
ordered by increasing longitude
observed of 26 mm (ALIC). No clear geographical dependency (longitude, latitude,
and altitude) of the bias or the standard deviation could be observed in this
comparison.
The comparison between the 4 RT-PPP solutions and the ROB’s PPP postprocessing reference solution showed very similar results (Figs. 3.48 and 3.49),
with typical biases ranging from 5 to 16 mm, typical standard deviations ranging
from 4 to 7 mm (i.e. slightly better than when compared to the IGS final product),
and correlation coefficient above 0.99. One can also remark that the RT-PPP
solutions performs very well and homogenously in Belgium: all stations in the
middle of the graphs (i.e. from VEUR to VITH) have lower biases ranging from
1 to 4 mm. We can also note the slightly better global performance obtained with
the CNS91 orbit and clock product. This lower standard deviation can be due to a
faster convergence time for the solution using CNS91 in GPS + GLONASS mode,
but this still needs to be confirmed. Finally, processing jointly GPS + GLONASS
observations (ROBC, IGS03-GLO) provides consistent solutions as processing
GPS-only observations (ROBB, IGS03), in general at the 1–2 mm level.
112
J. Douša et al.
In conclusion, ROB started in June 2014 to compute operationally RT-PPP
products as a demonstration campaign using the G-Nut/Tefnut software from
GOP. This campaign has run now for more than 3 years. Results obtained so far
are very promising but leave space for further studies and improvements. This
RT-PPP production and their related developments will continue in the framework
of the IAG WG 4.3.7 ‘Real-time GNSS tropospheric products’, with the aim on a
more longer term to be operationally provided to E-GVAP. As the next natural step
in the developments, ROB participates in a case study that aims to simulate the
RT-PPP processing in off-line mode. The goals of this case study are explained in
the next paragraph.
3.4.4.2
Assessment of RT Products in Simulated RT Analysis
of Benchmark Data
As we said, the RT-PPP demonstration campaign provided promising results but
also leaved space for specific studies and improvements. Such kind of studies, like
studying the optimal processing settings according to e.g. weather type (severe
versus normal condition), studying the influence of the station equipment, or optimizing the convergence period etc. requires to be able to replay the processing
several times on the same dataset by changing solely one processing parameter at a
time. In other words: simulating the real-time processing in an offline mode for finetunings and assessments. One perfect candidate for this is the WG1 Benchmark
campaign, and at the end of the COST Action, we started the real-time benchmark
campaign with the following main objectives:
• Processing (a core group of) stations of the Benchmark campaign in simulated
real-time PPP mode with various real-time orbits and clocks products to produce
a final assessment of the capability of each real-time orbit and clock product.
• Develop optimal strategies for the estimation of ZTDs and tropospheric gradients
at high (time) resolution (e.g. 5 min), and assess them to reference solutions
and/or ZTD/GRD from NWM.
• Investigate the capability of dynamically constrain ZTDs and GRDs according to
the weather conditions (calm, moderate, turbulent, severe, etc.). This is very
important for natural hazard warnings.
• Produce a dataset of simulated real-time tropospheric products (ZTD, horizontal
gradients, SPD) using the final fine-tuning of all previous steps that can be re-used
to assess the capability/performances of these products in real studies/applications (e.g. a nowcasting case of severe weather).
• Production of IWV in almost real-time for non-numerical nowcasting
applications.
• Standardize methods and format, and provide guidelines towards operational
production of real-time tropospheric products (link to E-GVAP).
In that context, ROB processed the complete WG1 benchmark dataset in simulated real-time offline mode with various processing options and various orbit and
3 Advanced GNSS Processing Techniques (Working Group 1)
113
Orbit and Clocks Products
ZTD Constraints
GRD Constraints
IGS01
From 0.5 (tight) to 5.0 (loose)
From 0.05 (tight) to 0.4 (loose)
IGS02
By Step of 0.5
By step of 0.05
IGS03 (GPS-only)
Fig. 3.50 The different processing configuration tested by ROB in the context of the real-time PPP
benchmark campaign (simulated real-time PPP offline processing). It includes 4 orbit and clock
products configurations (no offline version of the CNS91 product could be found), 10 ZTD
constraints setup, and 8 horizontal gradients constraints setup
clock products, focusing on providing the final assessment of the orbit and clock
products, on testing different ZTD & GRD (dynamical/optimal) constrains, on
studying the dependency of the performance w.r.t. to the equipment (see text of
the RT demonstration campaign), and on convergence periods. Similarly as for the
regular assessment of the real-time demonstration campaign, we also reprocessed the
benchmark campaign with the BSW52 to produce final reference tropospheric
products both in PPP and in double-difference approach. In total, 320 flavors of
the RT-PPP products are available for inter-comparison (Fig. 3.50) and assessments.
These results have been produced towards the end of the COST Action but will be
analyzed in the context of the IAG WG 4.3.7.
3.4.5
GFZ Real-Time Product Development and Assessment
in RT Analysis
C. Lu
GFZ German Research Centre for Geosciences, Potsdam, Germany
e-mail: [email protected]
X. Li
GFZ German Research Centre for Geosciences, Potsdam, Germany
e-mail: [email protected]
The multi-constellation Global Navigation Satellite Systems (GNSS) offers promising potential for the retrieval of real-time (RT) atmospheric parameters to support
time-critical meteorological applications, such as nowcasting or regional short-term
forecasts. In this study, we processed GNSS data from the globally distributed MultiGNSS Experiment (MGEX) network of about 30 ground stations by using the
precise point positioning (PPP) technique for retrieving RT multi-GNSS tropospheric delays. RT satellite orbit and clock product streams from the International
GNSS Service (IGS) were used. Meanwhile, we assessed the quality of clock and
orbit products provided by different IGS RTS ACs, called CLK01, CLK81, CLK92,
GFZC2, and GFZD2, respectively. Using the RT orbit and clock products, the
114
J. Douša et al.
performances of the RT ZTD retrieved from single-system as well as from multiGNSS combined observations are evaluated by comparing with the U.S. Naval
Observatory (USNO) final troposphere products. With the addition of multi-GNSS
observations, RT ZTD estimates with higher accuracy and enhanced reliability can
be obtained compared to the single-system solution. Comparing with the GPS-only
solution, the improvements in the initialization time of ZTD estimates are about
5.8% and 8.1% with the dual-system and the four-system combinations, respectively. The RT ZTD estimates retrieved with the GFZC2 products outperform those
derived from the other IGS RTS products, Fig. 3.51. In the GFZC2 solution, the
accuracy of about 5.05 mm for the RT estimated ZTD can be achieved with fixing
station coordinates. The results also confirm that the accuracy improvement (about
22.2%) can be achieved for the real-time estimated ZTDs by using multi-GNSS
observables compared to the GPS-only solution. In the multi-GNSS solution, the
accuracy of real-time retrieved ZTDs can be improved by a factor of up to 2.7 in the
fixing coordinate mode comparing with that in the kinematic mode, Fig. 3.52.
3.4.6
Contribution to RT Demonstration Campaign from
ULX9
W. Ding
University of Luxembourg, Luxembourg, Luxembourg
e-mail: [email protected]
F. N. Teferle
University of Luxembourg, Luxembourg, Luxembourg
e-mail: [email protected]
After initial RT solutions based on BNC were abandoned, ULX modified the
PPP-Wizard developed by CNES to provide a real-time solution to the
RT-Demonstration Campaign. These solutions included GPS, GLO and GAL observations and employed the real-time products from CNES CLK93, including satellite
orbit, clock and code/phase biases. The latter allowed PPP ambiguity resolution for
GPS using a zero-difference ambiguity resolution approach (see Table 3.16). The
modifications of PPP-Wizard included:
•
•
•
•
•
•
•
9
Apply Antenna Reference Point (ARP) correction from igs08.atx
Apply receiver PCO + PCV correction from igs08.atx
Solid earth tide + ocean tide loading correction (FES2004)
ZTD (GPT and Saastamoinen) + ZWD (modeled as random walk process)
Troposphere Mapping Function (GMF)
Elevation dependent weighting strategy (Q ¼ 1/cos(zen)**2)
Of particular interest was the impact of multi-GNSS on initialization times.
Parts from this section were previously published in Ding et al. (2017).
3 Advanced GNSS Processing Techniques (Working Group 1)
115
50
40
30
20
10
0
CLK01
Along ( cm )
1
2
3
5
7
6
8
50
40
30
20
10
0
1
2
C02
C03
4
3
5
6
C06
CLK92
GFZC2
GFZD2
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
GPS PRN
7
8
C08
C09
9
10
50
40
30
20
10
0
C05
CLK81
C07
12 13 14 15
GLONASS PRN
C11
C10
C12
C13
17
16
E11
C14
18
E12
19
E19
20
E22
21
22
E24
23
E26
24
E30
BDS+GALILEO PRN
50
40
30
20
10
0
Cross ( cm )
1
2
3
5
7
6
8
50
40
30
20
10
0
1
2
C02
C03
4
3
5
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
GPS PRN
6
7
8
C08
C09
9
10
50
40
30
20
10
0
C05
C06
C07
12 13 14 15
GLONASS PRN
C11
C10
C12
C13
17
16
E11
C14
18
E12
19
E19
20
E22
21
22
E24
23
E26
24
E30
BDS+GALILEO PRN
20
15
10
5
0
Radial(cm)
1
2
3
5
6
7
8
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
GPS
20
15
10
5
0
1
2
3
4
5
6
7
8
9
10
12
13
14
15
16
17
18
19
E12
E19
20
21
22
23
24
GLONASS
50
40
30
20
10
0
C02
C03
C05
C06
C07
C08
C09
C10
C11
C12
C13
C14
E11
E22
E24
E26
E30
BDS+GALILEO
Fig. 3.51 RMS values of the differences between IGS RTS orbits (CLK01, CLK81, CLK92,
GFZC2, and GFZD2) and GFZ final orbits for the four systems (i.e., GPS, GLONASS, Galileo, and
BDS) in the along (top), cross (middle) and radial (bottom) components, respectively
116
J. Douša et al.
CLK01
CLK81
CLK92
GFZC2
GFZD2
GPS RT ZTD (mm)
2600
2500
2400
2300
10min
20min
30min
40min
50min
60min
10min
20min
30min
40min
50min
60min
10min
20min
30min
40min
50min
60min
10min
20min
30min
40min
50min
60min
10min
20min
30min
40min
50min
60min
10min
20min
30min
40min
50min
60min
G/R RT ZTD (mm)
2600
2500
2400
2300
G/R/E/C RT ZTD (mm)
2600
2500
2400
2300
Fig. 3.52 RT ZTD estimates at station ONS1 derived from the GPS-only (“GPS”, top), the
combined GPS/GLONASS (“G/R”, middle), and the combined GPS/GLONASS/Galileo/BDS
(“G/R/E/C”, bottom) solutions in fixing coordinate (left panels) and kinematic processing (right
panels) modes by employing different IGS RTS over the first 2 h of DOY 090, 2017
Table 3.16 List of data processing modes investigated at ULX
Modes
RFLT
GFLT
GFIX
MFLT
MFIX
Details
Float PPP solution based on GLONASS-only observations
Float PPP solution based on GPS-only observations
Fixed PPP solution based on GPS-only observations
Float PPP solution based on GPS/GLONASS observations
Fixed PPP solution based on GPS/GLONASS/Galileo observations
Reproduced from Ding et al. (2017)
3 Advanced GNSS Processing Techniques (Working Group 1)
117
Fig. 3.53 Average initialization time in all data processing modes. (Reproduced from Ding et al.
2017)
From Fig. 3.53 it can be found that the initialization time required by the RFLT
solution is still the longest. It exceeds 30 min for two stations. Compared with that,
the initialization time for the GFLT solution is shorter for all stations. The average
value is 613 s (approx. 10.2 min). By applying ambiguity resolution, the initialization time becomes shorter for most of the stations, and is 583.6 s (approx. 9.7 min)
on average. The initialization process can also be accelerated by utilizing GNSS
observations, for which it can be achieved on average in 533 s (approx. 8.9 min).
Again, this suggests that the effect of the observation geometry is larger than that of
ambiguity resolution in accelerating the initialization process, especially considering
that an initialization time is required to achieve the first ambiguity resolution. When
applying all three GNSS and ambiguity resolution in the MFIX solution, the
initialization process is finished on average in 508.3 s (approx. 8.5 min), and there
are only small differences between different stations, which leads to the highest
consistency in the solutions and reveals the benefit of GNSS observation and
ambiguity resolution for severe weather event monitoring (for more details see
Ding et al. 2017).
Furthermore, the solutions from the five different RT data processing modes can
be compared to the benchmark troposphere products from CODE and USNO, both
being GPS + GLONASS solutions (Table 3.17).
The statistics with respect to the two types of benchmark products are similar,
which further validates the reliability and consistency of the reference products. The
RMS of the RFLT solution for nearly all stations is smaller than 15 mm, and the
average value is about 11.16 and 13.98 mm with respect to the final troposphere
products from CODE and USNO. Compared with that, the RMS of the GFLT
solution is better. The RMS of all stations is better than 12 mm except MOBS,
and is about 9 mm on average. The worse performance of the RFLT solution may
118
J. Douša et al.
Table 3.17 Mean accuracy of processing modes with respect to final troposphere products from
CODE and USNO (reproduced from Ding et al. 2017)
RFLT
GFLT
GFIX
MFLT
MFIX
CODE
Mean(mm)
0.82
0.83
2.09
0.47
1.48
STD(mm)
11.26
6.32
5.65
6.41
5.96
RMS(mm)
11.61
7.05
6.37
6.87
6.42
USNO
Mean(mm)
0.61
0.59
2.03
0.41
1.52
STD(mm)
13.67
8.27
7.45
8.27
7.69
RMS(mm)
13.98
8.95
8.17
8.69
8.14
come from two points: (1) the number of GLONASS satellites is less than for GPS;
(2) the accuracy of satellite products for GLONASS is worse than for GPS (Dach
and Jean 2015). However, considering the accuracy requirements (10–15 mm) in
updating NWP models, the RT troposphere estimates based on GPS or GLONASS
only observations can both fulfill the requirements (De Haan 2006).
Applying ambiguity resolution, the GFIX solution is further improved up to
0.8 mm on average compared to the GFLT solution. However, the mean bias
becomes slightly bigger. Combining the observations of two systems, the MFLT
solution is only 0.18 mm and 0.26 mm improved on average with respect to CODE
and USNO products, which reveals that the accuracy is not greatly improved by
incorporating GLONASS observations. In addition, the accuracy even becomes a
little worse for some stations, which may be correlated with the weighting strategy
between two systems and needs further research in the future. At last, the mean RMS
of the MFIX solution is 6.42 mm and 8.14 mm with respect to CODE and UNSO
products, respectively. It is the best solution among all the data processing modes,
which again reveals the effect in utilizing GNSS observations and ambiguity
resolution.
A further comparison was performed with respect to radiosonde observations
(Fig. 3.54). To summarize the accuracy of the 13 stations, we sort the results
according to the distance between the GNSS station and the nearby radiosonde
launch site. Since the mean bias of RT ZTD are monitored and will be corrected in
the assimilation procedure, we will only calculate the standard deviation (StDev) of
all stations (Bennitt and Jupp 2012). Based on the results, the accuracy of the RFLT
solution is the worst, of which the StDev is especially larger and exceeds 15 mm in
several stations. Among the other solutions, the StDevs are all smaller than 15 mm
except for ABMF and JFNG. On average, the StDevs of the two single system
solutions are 14.6 mm and 9.1 mm each, which again reveals that they can fulfill the
requirements in monitoring severe weather events.
However, compared with the GFLT solutions, we notice that the accuracies of the
GFIX and MFLT solutions become a little lower for many stations. Since there is
only one radiosonde observation in each day, this might be a consequence of the
instability of GPS phase bias information and the satellite orbit/clock products for
GLONASS. Additional details can be found in Ding et al. 2017.
3 Advanced GNSS Processing Techniques (Working Group 1)
119
Fig. 3.54 Standard deviation of RT ZTD errors with respect to the radiosonde observations in all
data processing modes (reproduced from Ding et al. 2017)
3.4.7
New Adaptable Strategy for RT and NRT Troposphere
Monitoring10
P. Václavovic
Geodetic Observatory Pecný, Research Institute of Geodesy, Topography and
Cartography, Zdiby, Czech Republic
e-mail: [email protected]
J. Douša
Geodetic Observatory Pecný, RIGTC, Ondřejov, Czech Republic
e-mail: [email protected]
In order to optimize the accuracy and the timeliness of tropospheric parameters with
possibly prioritizing the product latency in real-time (RT) or its accuracy in near realtime (NRT), we have developed a new adaptable strategy in the G-Nut/Tefnut
software (Douša et al. 2018a) exploiting the Precise Point Positioning, PPP
(Zumberge et al. 1997). We aimed to support various users from a single processing
PPP engine combining RT and NRT analysis modes. All the parameters, ZTD, GRD
and STD, are estimated consistently for both product lines when supported by realtime orbit and clock products. The strategy based on data from an individual station
only is also ready to exploit optimally all available multi-GNSS observations when
supported by precise products.
10
Parts from this section were previously published in Douša et al. (2018a).
120
3.4.7.1
J. Douša et al.
Epoch-Wise Filtering vs. Batch Processing
Nowadays, a common procedure of the NRT analysis within the EIG EUMETNET
GNSS Water Vapour Programme (E-GVAP, http://egvap.dmi.dk) is based on the
least-squares adjustment (LSQ) analysis and a piece-wise linear function of the
modelling of tropospheric parameters within the processing interval. By using the
BSW52 (Dach et al. 2015), Geodetic Observatory Pecný (GOP) applies the same
strategy in all long-term contributions to E-GVAP (Douša 2001a, b, Douša and
Bennitt 2013). The estimated tropospheric parameters for such case are displayed in
Fig. 3.55 by black dots and the piece-wise deterministic model by dash lines
connecting the parameters within each hourly product update; not necessarily
connected at update boundaries. According to the E-GVAP conventions, the last
product value is shifted by 1 min (HR:59) in order to avoid a duplicate value with the
next hour product update.
So far, the tropospheric horizontal linear gradients and slant delays were not
provided by GOP because of two reasons: (1) not being yet assimilated into
numerical weather models (NWM), and (2) significantly increase the number of
parameters in the network solution and, consequently, the computation time. From
this point of view, the epoch-wise processing (e.g. Kalman filter) is an optimal
strategy as it, contrary to the LSQ, estimates recurrently all unknown parameters in
every epoch. One of the consequences is that only previous observations contribute
to a current estimate and thus the solution needs a certain time to converge.
However, an accuracy of parameters estimated during the initial convergence, or
any later re-convergence, can be improved only by using the backward smoothing
algorithm (Václavovic and Douša 2015) additionally using both past and following
observations for improving the precision at every epoch. As a consequence, the
backward smoothing was designed for the post-processing solutions and it can
substitute the LSQ in a number of applications. In the first step of the filter,
parameters are predicted via adding particular amount of noise to diagonal elements
of the variance-covariance matrix belonging to dynamic parameters. When new
observations are available the state vector with its variance-covariance matrix are
Fig. 3.55 Various strategies of the troposphere modelling: (1) piece-wise model (black dots
connected with dash lines) and (2) stochastic modelling by real-time Kalman filter (white points),
hourly backward smoothing (blue points), and 45-min postponed hourly backward smoothing (red
points)
3 Advanced GNSS Processing Techniques (Working Group 1)
121
updated. Results from the prediction as well as from the update needs to be stored for
future smoothing. Since the standard Kalman filter and also smoother can suffer
from numerical instabilities due to the round-off error we have implemented alternative forms of the algorithms exploiting Cholesky and Singular Value Decompositions of the variance-covariance matrix instead of the original one.
3.4.7.2
RT and NRT Combined Processing Supported by Observations
from Files or Streams
In E-GVAP, the NRT products are updated on hourly basis and the delivery
requirement to the E-GVAP server at UK Met Office is 45 min after the last
observation used in the analysis. The NRT LSQ batch processing, initiated every
hour when obtaining a majority of data files, is thus a relevant solution for this
purpose when using hourly data files. Although the backward smoothing approach is
the most beneficial for the post-processing solutions, it can be effectively used in
NRT applications too. We have thus used it for the new adaptable strategy combining a continuously running forward filter with a regularly triggered backward
smoothing filter. The former is aimed for the estimating epoch-wise tropospheric
parameters in real-time indicated by white points in Fig. 3.55. The backward
smoothing is started periodically at a pre-defined time stamp as indicated at HR:00
in the Figure. It uses the initial state vector from the same epoch provided by the realtime filter for recalculating the past parameters from the Kalman filter. Obviously,
such recalculation is able to refine significantly older parameters, but cannot improve
parameters at the initial epochs of the backward smoothing. The length of the
smoothing period can be set flexibly to reflect an actual user preference for a higher
accuracy or a shorter latency of the product. In such way, the standard E-GVAP NRT
tropospheric product can also be provided on hourly basis as shown by blue points in
the Figure.
Initially, the new adaptable strategy was designed to use the precise orbit and
clock corrections disseminated through RT streams, it can however exploit observations coming from both real-time (streams) and near real-time (hourly or
sub-hourly files). While the former is necessarily used in a simultaneous RT and
NRT product generation, the latter is applicable for NRT only. In any case, both data
flows can be mixed and analysed for each station independently. Additional advantage of the NRT analysis utilizing the backward smoothing and RT observations
may profit from the 45-min requirement in E-GVAP for the product delivery in
NRT. By starting the backward smoothing shortly before the delivery request,
indicated in Fig. 3.55 by the red arrow starting at 12:40, new observations can
further improve the accuracy of NRT product, in particular last parameters of the
NRT product and still reduce systematic errors typical for data interval boundaries.
122
3.4.7.3
J. Douša et al.
Estimating High-Resolution ZTDs and Horizontal Gradients
Estimating high-resolution parameters with the batch LSQ can be more difficult due
to increasing size of normal equations, in particular for the network processing, and
consequently extending the processing time due to the inversion of the large matrix
of the solution. On the contrary, the described combination of the Kalman filter and
smother seems to be more convenient for monitoring dynamical processes when
high-resolution parameters are required. Since only observations from past epochs
are used in the Kalman filter processing, achieved parameters cannot react on their
fast change, and batch processing can reach better precision in this case. However,
when backward smoothing is applied, following observations improve the state
vector estimation when preserving high resolution from the previous forward filter.
Such situation is presented in Fig. 3.56 showing the ZTDs from different solutions during the fast change in the troposphere, indicated with a sudden decrease of
ZTD by 6–7 cm during 2.5 h at POTS station on May 10, 2013. Obviously, the postprocessing 15-min ZTDs from the GFZ solution using the EPOS software (Gendt
et al. 2004) implementing the LSQ processing (gray points) and the daily smoothed
30-s stochastic ZTDs from our software (black dots) are in a very good agreement.
Note that ZTDs from our solution is resampled to 5 min in the Figure. Due to the use
of past observations only, the real-time Kalman filter (red points) shows a delay in
the change of estimated parameter. The random walk for ZTD was set to 5 mm/h1/2.
The Figure reveals a characteristic behaviour for the real-time processing when using
past observations in stochastic parameter estimation. We can observe significant
improvements mainly in reducing the systematic behaviour for the 1-h backward
smoothing (green points). The main improvement can be reached within 15–30 min,
as indicated for the ZTD from the backward smoothing at 12:00, 13:00 and 14:00.
Interestingly, the ZTD from the 2-h smoothing already shows a very good agreement
with the post-processing LSQ solution. It shows a potential improvement discussed
in the previous subsections and postponing the NRT smoothing by at least 30 min if
RT observations are available.
Fig. 3.56 Real-time (red), near real-time (green, blue) and reference (black) ZTD estimates during
a fast change in the troposphere
3 Advanced GNSS Processing Techniques (Working Group 1)
3.4.7.4
123
Assessment of New Method Compared to the Existing E-GVAP
Processing
The new strategy has been initially developed and assessed using the GNSS4SWEC
Benchmark campaign (Douša et al. 2016) and firstly compared to the GOP NRT
tropospheric solution contributing operationally to E-GVAP. Thought the new
strategy can provide RT and NRT products in high temporal resolution, we compared only the ZTD as a product of HH:00 and HH:59 time stamps in every hour
representing the standard NRT E-GVAP product, see Fig. 3.55.
Table 3.18 summarizes results of three strategies and six ZTD solutions using
13 EUREF stations selected from the benchmark campaign and exploiting the
EUREF combined ZTD product as a reference for all comparisons (Pacione et al.
2017). The table shows summary statistics indicating similar improvements in terms
of the standard deviation over all ZTDs estimated at HH:00 in NRT independently
on applied products (IGS RTS vs. IGS final products), processing strategies and
software (G-Nut/Tefnut PPP vs. BSW52 DD). Compared to the Kalman filter ZTD
estimated at the last epoch (HH:59), the backward smoothing running on hourly
basis showed the improvement of 20% and 24% for the IGS RTS and the IGS final
product, respectively. The E-GVAP/GOP product demonstrates a similar improvement (24%) comparing ZTD from HH:00 against HH:59, which corresponds to our
previous results (Douša and Souček 2005). The new adaptable PPP solution using
the IGS final orbits reached the same accuracy as the E-GVAP/GOP product using
the IGS ultra-rapid orbits and NRT DD network solution from the BSW. On the
other hand, the use of IGS RTS products instead of IGS final products in PPP
indicates a degradation of 18% in ZTD SDEV and a 2.5 mm bias. It should be finally
noted, that the E-GVAP/GOP solution and the reference EUREF solution are based
on a similar processing strategy and the software, while the new strategy is significantly different.
Table 3.18 Summary statistics of three processing strategies and six ZTD solutions compared to
EUREF combined tropospheric product
Solution
RT PPP
(HR:59)
NRT PPP
(HR:00)
PP PPP
(HR:59)
PP PPP
(HR:00)
NRT DD
(HR:59)
NRT DD
(HR:00)
Software
G-Nut/
Tefnut
G-Nut/
Tefnut
G-Nut/
Tefnut
G-Nut/
Tefnut
BSW52
BSW52
Strategy description
Kalman filter, simulated realtime solution
Hourly backward smoothing in
real-time
Kalman filter in offline
processing, IGS final
Hourly backward smoothing
with IGS final
Last ZTD of hourly PW linear
LSQ
First ZTD of hourly PW linear
LSQ
Latency
<5 min
~
60 min
< 5 min
~
60 min
~
90 min
~
30 min
Mean
BIAS
2.4 mm
Mean
SDEV
5.7 mm
2.5 mm
4.6 mm
0.1 mm
4.7 mm
0.2 mm
3.6 mm
0.4 mm
4.9 mm
0.2 mm
3.7 mm
124
J. Douša et al.
Fig. 3.57 ZTD standard deviations and biases for 13 EUREF stations and six processing strategies
compared to the reference EUREF product
Figure 3.57 shows standard deviations and biases individually for all stations
comparing the first ZTDs (HH:00) and the last ZTDs (HH:59). The statistics of
ZTDs from the E-GVAP/GOP solution (PPP:DD-ultra) are plotted in red and pink
for HH:00 and HH:59, respectively. The results from the new strategy using the PPP
with IGS final orbit and clock products (PPP:IGS-final) are shown in dark and light
blue and, using IGS RTS (PPP:IGS03) in black and grey. Standard deviations for all
the stations show a similar improvement in the ZTD SDEV over all the strategies,
software and precise products. However, there is no significant impact of the strategy
on systematic errors, and we can observe only a common positive bias attributed to
the use of the IGS RTS products.
3.4.8
Optimum Stochastic Modeling for GNSS Tropospheric
Delay Estimation in Real-Time11
T. Hadaś
Wrocław University of Environmental and Life Sciences, Wrocław, Poland
e-mail: [email protected]
K. Kaźmierski
Wrocław University of Environmental and Life Sciences, Wrocław, Poland
e-mail: [email protected]
P. Hordyniec
Wrocław University of Environmental and Life Sciences, Wrocław, Poland
e-mail: [email protected]
11
Parts from this section were previously published in Hadaś et al. (2017).
3 Advanced GNSS Processing Techniques (Working Group 1)
125
J. Bosy
Wrocław University of Environmental and Life Sciences, Wrocław, Poland
e-mail: [email protected]
F. N. Teferle
University of Luxembourg, Luxembourg, Luxembourg
e-mail: [email protected]
It is commonly accepted by the GNSS community to constrain epoch-wise ZWD
estimates, usually by estimating ZWD as a random walk parameter. Wrocław
University of Environmental and Life Sciences (WUELS) have shown that the
optimum ZWD constraints in real-time GNSS processing, modelled as a randomwalk process, should be time and location specific. A typical approach is to perform
an initial empirical testing in order to obtain the effective constraining. As an
alternative, WUELS proposed to take benefit from numerical weather prediction
models to define optimum random walk process noise (RWPN). Two different
strategies were proposed and validated.
In the first approach an archived ZTD time series can be used to calculate a grid of
yearly means of the difference of ZWD between two consecutive epochs divided by
the root square of the time lapsed, which can be considered as a random walk process
noise. Using archived VMF-G grids, we obtained RWPN global grids for hydrostatic
and wet parameter (Fig. 3.58). We noticed that grids are nearly identical year by
year, with differences below 1 mm/√h for hydrostatic and wet grids. This means that
a single RWPN grid can be implemented in a software as a look up table to define the
optimum wet RWPN value for any station located worldwide. It was shown that
RWPN values from grids are similar to those obtain with empirical testing (see
Fig. 3.58).
Alternatively, a short-term weather forecast can be used to perform ray-tracing in
order to obtain forecast of ZTD and then to calculate RWPN dynamically in realtime. This approach was validated using forecast from GFS4 model. In this case
Fig. 3.58 Hydrostatic (left) and wet (right) yearly mean RWPN grid for 2015
126
J. Douša et al.
Fig. 3.59 Comparison of wet RWPN, ZTD time series, standard deviations of real-time ZTD
residuals
superior results were obtained, by means on the accuracy and precision of estimated
tropospheric delay (see Fig. 3.59). The advantage of this approach is that the wet
RWPN is regularly adjusted to the current tropospheric conditions. Its value remains
low, when ZTD is stable over time, and rises when a rapid change of ZTD is
expected. More details can be found in Hadaś et al. (2017).
3.5
Exploiting NWM-Based Products for Precise
Positioning
Numerical Weather Model (NWM) data and/or derived climatologies are increasingly used in precise geodetic applications. A prominent example is the Vienna
Mapping Function, VMF (Boehm et al. 2006a), and the Global Mapping Function,
GMF, (Boehm et al. 2006b). In recent years, the tropospheric models became more
sophisticated including NWM tropospheric (1st/2nd order) gradient estimates,
NWM tropospheric mapping factors, advantages of utilization of high-resolution
NWM models and NWM predictions, and others. The use of NWM-derived parameters plays more significant role in both a long-term reprocessing and real-time
kinematic positioning. The following section provides an overview of achievements
in this respect within the COST Action.
3 Advanced GNSS Processing Techniques (Working Group 1)
3.5.1
127
Tropospheric Parameters from Numerical Weather
Models
F. Zus
GFZ German Research Centre for Geosciences, Potsdam, Germany
e-mail: zusfl[email protected]
The tropospheric delay T is approximated as
T ðe; aÞ ¼ mf h ðah; bh; ch; eÞ ∙ Z H þ mf w ðaw; bw; cw; eÞ ∙ Z W
þ mf G ðC; eÞ½N cos ðaÞ þ E sin ðaÞ
ð3:20Þ
where e and a denote the elevation and azimuth angle of the station satellite link, mfh
and mfw denote the hydrostatic and non-hydrostatic Mapping Function (MF), mfg
denotes the gradient MF, Zh and Zw are the zenith hydrostatic and non-hydrostatic
delay, and N and E denote the (first-order) gradient components. The elevation angle
dependency of the hydrostatic (non-hydrostatic) MF is based on the continued
fraction form proposed by Marini (1972) and normalized by Herring (1992) to
yield the unity at zenith
mða; b; c; eÞ ¼ ð1 þ a=ð1 þ b=ð1 þ cÞÞÞ=ð sin ðeÞ þ a=ð sin ðeÞ þ b=ð sin ðeÞ þ cÞÞÞ
ð3:21Þ
The elevation angle dependency of the gradient MF is based on the form
proposed by Chen and Herring (1997):
mgðC; eÞ ¼ 1=ð sin ðeÞ tan ðeÞ þ C Þ
ð3:22Þ
where C ¼ 0.003. Therefore, provided that the tropospheric parameters ah, bh, ch,
aw, bw, cw, zh, zw, N and E are known, the tropospheric delay can be assembled for
any station satellite link. The tropospheric parameters are determined from
ray-traced tropospheric delays. The required pressure, temperature and humidity
fields are taken from a Numerical Weather Model (NWM). At GFZ Potsdam the
algorithm proposed by Zus et al. (2014) is used to compute mapping factors and slant
factors, i.e. the ratios of slant and zenith delays. Note that mapping factors are
computed under the assumption of a spherically layered troposphere. From the
mapping factors and slant factors the tropospheric parameters are estimated by
least-squares fitting. This is done separately for the mapping function coefficients
and the gradient components.
128
3.5.1.1
J. Douša et al.
Mapping Function Coefficients
For each station, 10 hydrostatic (non-hydrostatic) mapping factors are computed for
elevation angles of 3, 5, 7, 10, 15, 20, 30, 50, 70, 90 degree, and the hydrostatic
(non-hydrostatic) MF coefficients are determined by least-squares fitting (Zus et al.
2015a).
3.5.1.2
Gradient Components
At first, 120 slant factors and corresponding mapping factors are computed at
elevation angles 3, 5, 7, 10, 15, 20, 30, 50, 70, 90 degree and the azimuth angles
0, 30, 60, 90, 120, 150, 180, 210, 240, 270, 300, 330 degree. Second, zenith delays
are applied to obtain azimuth-dependent and azimuth-independent slant total delays.
Third, the differences between azimuth-dependent and azimuth-independent slant
total delays are computed. Finally, the gradient components are determined by leastsquares fitting (Zus et al. 2015b).
3.5.1.3
Tropospheric Model Errors
The pressure, temperature and humidity fields are taken from Global Forecast
System (GFS) of the National Centers for Environmental Prediction (NCEP). The
NCEP’s GFS analyses are available every 6 h (00:00, 06:00, 12:00, 18:00 UTC) with
a horizontal resolution of 0.5 degree on 31 pressure levels. Tropospheric parameters
are derived for a grid with a resolution of 0.5 degree. For any gridpoint on Earh’s
surface we examine how well the modeled tropospheric delays, i.e., the tropospheric
delays assembled from the tropospheric parameters, match the true (ray-traced)
tropospheric delays. This is done by calculating the elevation angle dependent
postfit-residual.
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2
P
S e; a j
T e; a j
RðeÞ ¼
n
ð3:23Þ
where T denotes the assembled tropospheric delay and S denotes the ray-traced
tropospheric delay. The elevation angles are chosen to be 3, 5, 7, 10, 15, 20, 30, 50,
70, 90 degree and the azimuth angles are chosen to be 0, 30, 60, 90, 120, 150,
180, 210, 240, 270, 300, 330 degree. As an example we consider a single epoch (first
of May 2017, 12:00 UTC).
At first, we examine azimuth independent tropospheric delays. Figure 3.60 shows
the residuals for an elevation angle of 7 . The residuals are well below 1 mm for any
point on Earth’s surface. For comparison Figure 3.61 shows the residual for an
elevation angle of 7 if we replace the NWM based MF by the GMF (Boehm et al.
2006a, b). Clearly, as the GMF is based on a climatology it cannot capture the short-
3 Advanced GNSS Processing Techniques (Working Group 1)
129
MF rmse [mm] elevation 7 deg
2
80
1.8
60
1.6
latitude [deg]
40
1.4
20
1.2
1
0
0.8
–20
0.6
–40
0.4
–60
0.2
–80
50
100
150
200
250
longitude [deg]
300
350
0
Fig. 3.60 We examine azimuth independent tropospheric delays (1st of May 2017, 12:00 UTC).
The scatter plot shows the tropospheric residuals for an elevation angle of 7
GMF rmse [mm] elevation 7 deg
30
80
60
25
40
latitude [deg]
20
20
15
0
–20
10
–40
5
–60
–80
50
100
150
200
250
300
350
0
longitude [deg]
Fig. 3.61 We examine azimuth independent tropospheric delays (1st of May 2017, 12:00 UTC).
The NWM based MF is replaced by the GMF. The scatter plot shows the tropospheric residuals for
an elevation angle of 7
130
J. Douša et al.
50
MF
GMF
average rmse per station [mm]
45
40
35
30
25
20
15
10
5
0
3
5
7
10
elevation angle [deg]
15
Fig. 3.62 We examine azimuth independent tropospheric delays (1st of May 2017, 12:00 UTC).
The bar plot shows the average residual per grid point as a function of the elevation angle. Different
colors show different options; NWM based MF and GMF
term variability of the troposphere and hence the residuals are larger. Figure 3.62
shows the average residual per grid point as a function of the elevation angle. The
residuals for the NWM based MF are well below 1 mm for any elevation angle
emphasizing the fact that the three term continued fraction form of the mapping
function works with an exquisite level of precision. Hence, if the troposphere is
indeed spherically layered, the tropospheric model (its functional form) can be
regarded error free.
Next, we examine azimuth dependent tropospheric delays. Figure 3.63 shows the
residuals for an elevation angle of 7 when no gradients are applied and Fig. 3.64
shows the residuals for an elevation angle of 7 when gradients are applied. As to
expect the residuals are significantly reduced when gradients are applied. The
residuals can be reduced by adding higher-order gradients to the tropospheric
delay model. In essence, the tropospheric delay is approximated as
T ðe; aÞ ¼ mf h ðah; bh; ch; eÞ ∙ Z H þ mf w ðaw; bw; cw; eÞ ∙ Z W
þ mf G ðC; eÞ½GN cos ðaÞ þ GE sin ðaÞ
þ mf G ðC; eÞ½FcosðaÞ cos ðaÞ þ GcosðaÞ sin ðaÞ þ HsinðaÞ sin ðaÞ
ð3:24Þ
where F, G and H denote higher-order gradient components.
Figure 3.65 shows the residuals for an elevation angle of 7 when first- and
higher-order gradients are applied. Finally, Fig. 3.66 shows the average residual per
grid point as a function of the elevation angle. The application of higher-order
3 Advanced GNSS Processing Techniques (Working Group 1)
131
00 rmse [mm] elevation 7 deg
50
80
45
60
40
latitude [deg]
40
35
20
30
0
25
–20
20
15
–40
10
–60
5
–80
50
100
150
200
longitude [deg]
250
300
350
0
Fig. 3.63 We examine azimuth dependent tropospheric delays (1st of May 2017, 12:00 UTC). The
scatter plot shows the tropospheric residuals for an elevation angle of 7 when no gradients are
applied
01 rmse [mm] elevation 7 deg
50
80
45
60
40
latitude [deg]
40
35
20
30
0
25
–20
20
15
–40
10
–60
5
–80
50
100
150
200
longitude [deg]
250
300
350
0
Fig. 3.64 We examine azimuth dependent tropospheric delays (1st of May 2017, 12:00 UTC). The
scatter plot shows the tropospheric residuals for an elevation angle of 7 when first-order gradients
are applied
132
J. Douša et al.
01+02 rmse [mm] elevation 7 deg
50
80
45
60
40
latitude [deg]
40
35
20
30
0
25
–20
20
15
–40
10
–60
5
–80
50
100
150
200
250
longitude [deg]
300
0
350
Fig. 3.65 We examine azimuth dependent tropospheric delays (1st of May 2017, 12:00 UTC). The
scatter plot shows the tropospheric residuals for an elevation angle of 7 when first- and higherorder gradients are applied.
90
O1+O2
O1
O0
average rmse per station [mm]
80
70
60
50
40
30
20
10
0
3
5
7
10
elevation angle [deg]
15
Fig. 3.66 We examine azimuth independent tropospheric delays (1st of May 2017, 12:00 UTC).
The bar plot shows the average residual per grid point as a function of the elevation angle. Different
colors show different options; no gradients applied, first-order gradients applied and first- and
higher-order gradients applied.
3 Advanced GNSS Processing Techniques (Working Group 1)
133
gradients reduces the residual in particular for elevation angles below 10 . We recall
that the underlying NWM has a horizontal resolution of 0.5 degree. Small scale
tropospheric features cannot be represented by the low horizontal resolution. Larger
residuals are expected when the horizontal resolution of the underlying NWM
increases.
3.5.2
The Impact of Global and Regional Climatology
on the Performance of Tropospheric Blind Models
G. Möller
Department of Geodesy and Geoinformation, TU Wien, Wien, Austria
e-mail: [email protected]
J. Sammer
Department of Geodesy and Geoinformation, TU Wien, Wien, Austria
e-mail: [email protected]
For tropospheric delay modelling usually the concept of mapping functions is used:
ΔLðεÞ ¼ ΔLhz ∙ mf h ðεÞ þ ΔLwz ∙ mf w ðεÞ
ð3:25Þ
It describes the total slant delay ΔL at elevation angle ε as as the sum of a
hydrostatic and a wet component. Each component can be expressed as the product
of a zenith delay and the corresponding mapping function.
Empirical troposphere models like GPT2w (Böhm et al. 2015) can provide this
information for any user position and epoch. In particular, GPT2w is based on global
1 1 gridded values of tropospheric state parameters like surface pressure for
modelling the zenith hydrostatic delay or water vapour pressure, weighted mean
temperature and water vapour decrease factor for modelling of the zenith wet delay.
In addition, mapping coefficients are provided separately for the hydrostatic and the
wet mapping function, to further reduce the mapping error, especially below
15 degrees elevation angle (Möller et al. 2014).
Comparison with time series of the IGS reveal that GPT2w allows for modelling
of the zenith tropospheric delay with a bias of less than 1 mm and a RMS of about
3.6 cm, see (Böhm et al. 2015). However, to further improve the model performance
on regional level, the global 1 1 grid was replaced by a regional 0.2 0.3 grid.
Analogous to the global grid, on the regional climatological grid tropospheric
parameters are provided as mean, annual and semi-annual coefficients. The regional
coefficients were derived from 3 years of ALARO model data, as provided by the
Central Institute for Meteorology and Geodynamics (ZAMG), Austria on 18 pressure
levels with a temporal resolution of 3 h.
Figure 3.67 shows time series of pressure (p), temperature (T), temperature lapse
rate (dT), mean temperature (Tm), water vapour pressure (e) and water vapour
134
J. Douša et al.
Fig. 3.67 Meteorological parameters as obtained from GPT2w (blue) and GPT2w ALARO (red) at
GNSS station HART, Austria. Analysed period: May–Dec 2013
Fig. 3.68 Histogram of ZTD residuals (GNSS minus model) as obtained at 45 GNSS in Austria.
(Left) GNSS minus GPT2w, (right) GNSS minus GPT2w ALARO. Analysed period: May–
Dec 2013
decrease factor (Lambda), as obtained from both grids, exemplary for GNSS station
HART, Austria. The regional model (GPT2w ALARO) shows in general a more
distinct seasonal signal than the global GPT2w. Especially in temperature lapse rate
and water vapour decrease factor, significant differences can be observed.
In order to evaluate the impact on the delay modelling based on both grids, ZTD
time series were derived for about 45 GNSS sites in Austria and neighbouring
countries. Figure 3.68 shows the ZTD residuals for GPT2w and GPT2w ALARO
with respect to ZTDs derived from dual-frequency GNSS observations at GNSS site.
The regional grid helps to further reduce the ZTD bias at GNSS sites, which are
located north and south of the main Alpine ridge. Nevertheless, averaged over all
3 Advanced GNSS Processing Techniques (Working Group 1)
135
GNSS station and over the period May to December 2013, the mean bias (GPT2w:
2 mm and GPT2w ALARO: 3 mm) and standard deviation (GPT2w: 29 mm and
GPT2w ALARO: 30 mm) of GPT2w ALARO is very comparable to the global
GPT2w. In consequence, it is concluded that a higher spatial resolution does not lead
consequently to a better performance of the tropospheric blind model, even not in the
Alpine area. In order to further increase the performance of tropospheric blind
models in future other strategies have to be discovered.
3.5.3
Refined Discrete and Empirical Troposphere Mapping
Functions VMF3 and GPT312
D. Landskron
Department of Geodesy and Geoinformation, TU Wien, Wien, Austria
e-mail: [email protected]
J. Boehm
Department of Geodesy and Geoinformation, TU Wien, Wien, Austria
e-mail: [email protected]
The Vienna Mapping Functions 3 (VMF3) is a refinement of VMF1 with the aim of
even higher precision. It eliminates shortcomings in the empirical coefficients b and
c and is not only tuned for the specific elevation angle of 3 , but for the whole
elevation range through least-squares adjustments. The new mapping function
coefficients were determined on the basis of ray-traced delays of the ray-tracer
RADIATE (Hofmeister and Böhm 2017). Comparing modeled slant delays of
VMF3 and VMF1 with the underlying ray-traced delays proves the high quality of
VMF3, in particular at low elevation angles. In consequence, when requiring highest
precision, VMF3 is to be preferable to VMF1. For more details, the reader is referred
to Landskron and Böhm (2017).
Figure 3.69 shows the empirical (blind) troposphere model Global Pressure and
Temperature 3, GPT3 (Landskron and Böhm 2017) as a refinement of the model
Global Pressure and Temperature 2 wet, GPT2w (Böhm et al. 2015), with
re-calculated mapping function coefficients and empirical horizontal gradients,
available in a horizontal resolution of 5 5 and 1 1 . The meteorological
quantities remain unchanged. The empirical mapping factors from GPT3 are averaged from the VMF3 applying information from 2D ray-tracing through numerical
weather models of the ECMWF.
GPT3 is full consistent with VMF3 and can be used for GNSS as well as VLBI
analysis. It is unique in such a way as it provides the entire information which is
required in order to model a priori the troposphere delay dependent on elevation
angle and azimuth.
12
Parts from this section were previously published in Landskron and Böhm (2017).
136
J. Douša et al.
Fig. 3.69 Hydrostatic north gradient (left) and hydrostatic east gradient (right) from GPT3. While
the north gradient shows systematic features for the northern and southern hemisphere owing to the
atmospheric bulge, the east gradient is rather affected by the land-ocean distribution
3.5.4
The Impact of NWM Forecast Length on ZTDs13
J. Douša
Geodetic Observatory Pecný, RIGTC, Ondřejov, Czech Republic
e-mail: [email protected]
P. Václavovic
Geodetic Observatory Pecný, Research Institute of Geodesy, Topography and
Cartography, Zdiby, Czech Republic
e-mail: [email protected]
M. Eliaš
Geodetic Observatory Pecný, Research Institute of Geodesy, Topography and
Cartography, Zdiby, Czech Republic
e-mail: [email protected]
P. Krč
Czech Institute of Computer Science, Academy of Sciences, Praha, Czech Republic
e-mail: [email protected]
K. Eben
Czech Institute of Computer Science, Academy of Sciences, Praha, Czech Republic
e-mail: [email protected]
J. Resler
Czech Institute of Computer Science, Academy of Sciences, Praha, Czech Republic
e-mail: [email protected]
To study the impact of the GNSS tropospheric corrections on the NWP prediction
length, we calculated ZTDs for stations in the Benchmark campaign using the
G-Nut/Shu software and the Weather Research and Forecasting (WRF) models
operated routinely by the Institute of Computer Science, Academy of Sciences,
13
Parts from this section were previously published in Douša et al. (2015a, b).
3 Advanced GNSS Processing Techniques (Working Group 1)
137
Czech Republic (ICS ASCR). The two same WRF-ICS analyses have routinely
contributed to the Real-time Demonstration campaign since July 2015 and can be
characterized as follows: (a) two regional domains 9 9 km (EU9) and 3 3 km
(CZ3), (b) uniform horizontal grid represented using the Lambert Conformal Conic
projection (LCC), (c) grid unstaggered dimensions (west-east ¼ 418 grid points;
south-north ¼ 302 grid points), (d) 38 vertical levels with the top level at 50 hPa,
(e) four forecasts per day: 00:00, 06:00, 12:00 and 18:00 UTC, (f) a 1-h temporal
resolution in each forecast, and (g) a 13-h length of the forecast.
The data used for the WRF-ICS analysis computation are collected within a 6-h
window surrounding the synoptic time T. The analysis then emerges on the web site
of NCEP about 3 h 25 min after the synoptic time. The global model runs the
forecasts for increasing time horizons up to 14 days. The resulted forecasts are
successively uploaded until about T + 5 h. The mesoscale WRF model starts its
simulation run at the synoptic time T, using previous analysis (T-6 h) and the time
window from T-6 h to T. The so called grid nudging is performed at this level. In this
way the mesoscale model reaches a state in time T, being a downscaled analysis.
Grid nudging ensures that the mesoscale model doesn’t diverge too far from the
global analysis in time T. The mesoscale simulation starts at approximately T + 3 h
35 min real time and during the grid nudging phase and the subsequent spinup phase,
i.e. while simulating hours (T, T + 6), the mesoscale model simulation catches up the
real time, until about T + 4 h 30 min real time. About T + 7 h real time when
depending of computing resources, the forecast horizon T + 78 h is produced.
Based on the WRF-ICS different forecasting intervals counted from the time T:
0-6 h (spinup), 6-12 h (forecast) and 12–18 h (forecast), we calculated ZTDs for all
Benchmark stations during May/June 2013 and compared then with reference GNSS
ZTD parameters. Figure 3.70 shows the comparisons of the NWP-based ZTDs with
Fig. 3.70 ZTD biases (top) and standard deviations (bottom) calculated for 420 stations of the
GNSS4SWEC Benchmark campaign over 2 weeks in May–June 2013 NWP models compared to
GOP GNSS reference solution. From left to right three prediction intervals are shown: 0–6 h (left),
6–12 h (middle) and 12–18 (right)
138
J. Douša et al.
Table 3.19 Summary statistics for the prediction length
NWP domain
D01/EUR
D01/EUR
D01/EUR
D02/CZ
D02/CZ
D02/CZ
Forecast window
0–6 h
6–12 h
12–18 h
0–6 h
6–12 h
12–18 h
Bias [mm]
1.50
0.89
0.51
+2.05
+2.46
+3.20
Sdev [mm]
9.93
10.95
12.91
9.14
10.33
12.83
RMS [mm]
10.42
11.55
13.48
9.84
11.90
13.50
respect to GNSS reference solution in geographical plots. Biases (top) and standard
deviations (bottom) are shown for the three prediction windows (left to right).
Table 3.19 then provides a summary statistics for ZTDs calculated from two WRF
domains. We focused on assessing zenith tropospheric delays potentially usable as
external corrections for GNSS real-time applications such as positioning and navigation. Statistics for 14 days and 420 stations of the GNSS4SWEC Benchmark
campaign [6] summarize ZTDs calculated from the spinup (0–6 h) and forecast
intervals (6–12 h and 12–18 h). The summary shows that the quality of ZTD within
the prediction windows up to 18-h resulted in RMS of 9.5–13.5 mm. It demonstrated
a slow degradation of ZTDs from NWM approximately at a rate of 1–2%/h, which
can be characterized by 1–2 mm/h. A degradation of mean biases was not observed.
ZTDs usable in real-time GNSS applications correspond to the prediction of 6–12 h
in the standard WRF operation.
3.5.5
Dual-Layer Tropospheric Correction Model Combining
Data from GNSS and NWM14
J. Douša
Geodetic Observatory Pecný, RIGTC, Ondřejov, Czech Republic
e-mail: [email protected]
M. Eliaš
Geodetic Observatory Pecný, Research Institute of Geodesy, Topography and
Cartography, Zdiby, Czech Republic
e-mail: [email protected]
P. Václavovic
Geodetic Observatory Pecný, Research Institute of Geodesy, Topography and
Cartography, Zdiby, Czech Republic
e-mail: [email protected]
14
Parts from this section were previously published in Douša et al. 2018b.
3 Advanced GNSS Processing Techniques (Working Group 1)
139
K. Eben
Czech Institute of Computer Science, Academy of Sciences, Praha, Czech Republic
e-mail: [email protected]
Various tropospheric models for GNSS real-time positioning have been developed at
GOP recently taking advantages of recent enhancements in the troposphere modelling (1) analytical ZWD calculation based on the concept of Askne and Nordius
(1987) when combining exponential decay parameters from the water vapor pressure
vertical profile and ZWD profile (Douša and Eliaš 2014), (2) more precise vertical
approximation for the ZWD parameter using new parameter for modelling of ZWD
exponential decay expressed either for a dependency on pressure or altitude (Douša
and Eliaš 2014), (3) more simple and accurate tropospheric model based on user
parameters related to the hydrostatic and non-hydrostatic zenith delays (Douša et al.
2015a, b), and (4) a flexible parameterization of new and legacy ZWD modelling
approaches in various user modes, (Douša et al. 2015a, b).
Based on the abovementioned enhancements in the modelling of tropospheric
corrections, GOP has developed a new concept of a dual-layer tropospheric correction model for GNSS precise real-time positioning applications which optimally
benefit from the synergy between NWM and GNSS data. The idea behind is to
combine and predict optimally hydrostatic and wet components of the total zenith
path delay in support of real-time GNSS positioning applications. In a simple form it
resembles the approach of an assimilation of GNSS ZTD into the numerical weather
forecast, however, targeting the GNSS user parameters, i.e. tropospheric path delays
of an electromagnetic signal up to 15GHz frequencies.
The new concept aimed at enhancing the original GOP augmentation model
introduced in Douša et al. (2015a, b) by combining NWM data with ZTDs estimated
from GNSS permanent stations in regional networks. The first layer is represented
with the background NWM-driven model available anytime for the region of interest
when derived purely from a NWM forecast provided from several hours up to
1–2 days. The first-layer parameters are predicted from NWM data fields and provides the background ZHD, ZWD together with auxiliary parameters for the parameter vertical scaling. The second layer improves mainly the ZWD, or more precisely,
optimized corrections complementary to the NWM-derived hydrostatic corrections
of the tropospheric effect on the at radio-frequency electromagnetic signal.
The description of the combination method is given in the paper by Douša et al.
(2017) and includes four test cases for optimal method of ZWD combination from
GNSS and NWM. Two NWM models were used: (1) global ERA-Interim reanalysis
(Dee et al. 2011) provided by ECMWF (we used 6-h updates with 1 1 horizontal
resolution), and (2) high-resolution regional NWM prediction (European model with
9 9 km horizontal uniform resolution, hourly updated) provided from an operational short-time prediction from the mesoscale WRF model operated by the Institute
of Computer Science (ICS), Academy of Sciences of the Czech Republic. The input
GNSS ZTDs stems from the analysis of GNSS data of the GNSS4SWEC WG1
Benchmark dataset (Douša et al. 2016) provided by GOP as the GNSS reference
solution.
140
J. Douša et al.
Fig. 3.71 Mean statistics of a single-layer model (top) and dual-layer (bottom) ZTDs with respect
to GNSS ZTDs using a closed-loop test case and ERA Interim (left) and WRF-ICS (right) models
during May, 2013
The concept was implemented in the G-Nut/Shu software for the testing and
assessing different variants of ZWD weighting method. The closed-loop results
demonstrated that the new combination concept provides highly accurate and stable
results despite of several involved approximations, see Fig. 3.71. The GNSS data
were able to improve significantly the non-hydrostatic part of the NWM-based
tropospheric model, however it is able to correct also possible errors in the hydrostatic component coming from the underlying NWM data. The most significant
improvement of using the second model layer was reached in term of ZTD standard
deviations, in total over 43%, when assessed using products from independent
GNSS stations. An improvement has also been observed in term of systematic
errors, but these were almost negligible in the total statistics. The most important
result of the combination of GNSS and NWM data was found in term of stability in
the session-to-session performance of the dual-layer model ZTD statistics. The
scatters in session-to-session mean standard deviations were reduced from 8.7 to
4.4 mm and from 9.6 to 3.0 mm for the WRF-ICS model when using GNSS ZTD and
ZTD together with horizontal gradients, respectively, see Table 3.20.
The GNSS contribution to the dual-layer augmentation model consists particularly in a local phenomena where the model resolution plays a key role for an optimal
assimilation of GNSS ZTDs into the model. We demonstrated a fast decrease of the
statistics when simulating a higher horizontal resolution for the ERA-Interim model
and obtaining comparable or better results to those resulting from the high-resolution
WRF-ICS model. Comparing several ZWD weighing methods in the combination,
optimal results were achieved by replacing the original NWM ZWDs with GNSS
ZWDs if these were interpolated robustly to the model grid points.
Using tropospheric horizontal gradients estimated for each GNSS station, we
developed a method to calculate so-called pseudo-ZTDs to densify GNSS ZTD field
for its optimal contribution to the combination with NWM ZWDs stemming from
the WRF-ICS high-resolution model, example is given in Fig. 3.72. The combination of GNSS ZTDs and pseudo-ZTDs with NWM showed up to 70% improvement
3 Advanced GNSS Processing Techniques (Working Group 1)
141
Table 3.20 ZTD mean statistics of GNSS and ERA-Interim data weighting within the ZWD
combination and improvements with respect to the background NWM model (last line)
Data reduction:
Variant
First layer (ERA)
First layer (WRF)
Second layer
(ERA)
Second layer
(WRF)
Data weighting [mm]
σ GNSS ¼ 10.0;
σ NWM ¼ 1
σ GNSS ¼ 10.0;
σ NWM ¼ 1
σ GNSS ¼ 1;
σ NWM ¼ 10.0
σ GNSS ¼ 1;
σ NWM ¼ 10.0
None
bias/sdev
[mm]
+2.6 / 9.3
R3 (33%)
bias/sdev
[mm]
+2.6 / 9.3
R2 (50%)
bias/sdev
[mm]
+2.7 / 9.4
0.8 / 9.6
0.7 / 9.5
0.8 / 9.6
+2.8 / 5.3
+2.7 / 5.2
+3.0 / 5.5
0.2 / 2.4
0.2 / 2.9
0.0 / 3.0
Fig. 3.72 Differences of ZTDs (points with circles) and pseudo-ZTDs at distances of 15 km and
25 km (small points) with respect to the reference GNSS ZTDs are showed for the model first layer
(top panels) and for the model second layer (bottom panels) and ERA-Interim (left panels) and
WRF-ICS (right panels), on May 31, 12:00 UTC, 2013
for standard deviations compared to the NWM model, and about 35% improvement
compared to the initial GNSS ZTDs and NWM data combined model. As expected
in this test case, we haven’t observed any improvement in case of the ERA-Interim
model.
As the developed method of NWM and GNSS data combination demonstrated a
high accuracy and stability over time, it is very promising for implementation of a
142
J. Douša et al.
service for tropospheric corrections for real-time precise positioning. Currently, we
are implementing the operational prototype combining the existing GOP services for
tropospheric parameters estimated (near) real-time (Douša and Václavovic 2014,
2016) provided in support of the numerical weather forecasting, and the existing
demonstration prototype for the GOP model background layer utilizing the
WRF-ICS operational weather forecast for Europe. More details can be found in
Douša et al. 2018a, b.
3.5.6
Tropospheric Refractivity and Zenith Path Delays from
Least-Squares Collocation of Meteorological and GNSS
Data15
K. Wilgan
Wrocław University of Environmental and Life Sciences, Wrocław, Poland
e-mail: [email protected]
W. Rohm
Institute of Geodesy and Geoinformatics, Wrocław University of Environmental and
Life Sciences, Wrocław, Poland
e-mail: [email protected]
J. Bosy
Wrocław University of Environmental and Life Sciences, Wrocław, Poland
e-mail: [email protected]
F. Hurter
ETH Zurich, Zürich, Switzerland
e-mail: [email protected]
A. Geiger
ETH Zurich, Zürich, Switzerland
e-mail: [email protected]
This subsection summarizes results of the troposphere model tested in two countries
with different orography: Switzerland (mountainous) and Poland (mostly flat). The
troposphere model is based on the least-squares collocation technique, where each
observation is divided into a deterministic part, a correlated stochastic part (signal)
and an uncorrelated stochastic part (noise). The selected parameters from different
data sources are estimated simultaneously in the least-squares sense taking into
account the two kinds of errors. The advantage of this method is a relatively easy
implementation of additional data sources. Using the computed model coefficients it
is possible to reconstruct the value of considered parameter at any time and place.
15
Parts from this section were previously published in Wilgan et al. (2017a)
3 Advanced GNSS Processing Techniques (Working Group 1)
143
Calculations were made using the software COMEDIE (Collocation of Meteorological Data for Estimation and Interpretation of tropospheric path delays), developed at
ETH Zürich.
For Switzerland, the profiles of total refractivity were calculated from three data
sets: (1) the total refractivity calculated from ground-based meteorological measurements, (2) the total refractivity calculated from meteorological measurements and
ZTD from GNSS stations and (3) the total refractivity calculated from meteorological measurements, ZTD GNSS and horizontal gradients of ZTD from GNSS
stations. The data set (2) exhibits the best agreement with the reference RS data.
The data set (1) based on the ground-based meteorological data also gives a good
information about the tropospheric state, but only up to the height of the highest
station. Above that height, it is necessary to use an additional data source, such as the
ZTD GNSS, to provide an information about tropospheric parameters in the vertical
profiles. Unfortunately, adding the horizontal gradients of ZTD in data set (3) did not
improve the troposphere model.
In the countries located mainly on lowlands such as Poland, the height distribution of ground-based meteorological stations is too flat to reconstruct the refractivity
profiles with the collocation technique. Thus, the troposphere model is built based on
NWP data of 10 km 10 km spatial resolution, 34 height levels and time resolution
of 1 h. The total refractivity profiles and ZTD values were calculated from 4 data
sets: (1) NWP WRF model, (2) WRF model integrated with ZTD GNSS, (3) ZTD
GNSS only, and (4) WRF model, ZTD GNSS and ground-based meteorological
measurements. The obtained total refractivity profiles were compared with the
reference radiosonde (RS) data and obtained ZTD with the reference GNSS data
from post-processing. For total refractivity, the best agreement with reference RS
data was achieved from the WRF model integrated with GNSS data (data set 2).
Including the ground-based meteorological data (data set 4) did not improve the
model. Data set (1) exhibited similar agreement with RS measurements as data set
(2), only in the upper layers of the model there was a displacement of 2 ppm. The
data set (3) showed much worse accuracy with the discrepancies at lower altitudes
even at the level of 30 ppm. Such large differences are a result of an attempt to
reconstruct a whole profile of refractivity from a single ZTD value. In the next step,
the model of ZTD from COMEDIE was compared with the reference near-real time
GNSS data. The comparisons were made for 9 days in May 2014 that included a
severe weather event. The data sets of the highest agreement with the reference
GNSS data are (2) and (3) with the average biases of 3.7 mm and 3.8 mm and
standard deviation of 16.7 mm and 17.2 mm, respectively. The collocation based
only on the WRF model (data set 1) overestimates the ZTD values. The reason for
such behavior is that the humidity values provided by the WRF model after the
rainfall are too high, which directly affects the ZTD values. To sum up, the
integration of NWP and GNSS data is essential to provide accurate models for
both total refractivity and ZTDs. More details can be found in Wilgan et al. 2017a, b.
144
3.5.7
J. Douša et al.
Improving Precise Point Positioning with Numerical
Weather Models
C. Lu
GFZ German Research Centre for Geosciences, Potsdam, Germany
e-mail: [email protected]
Precise positioning with the current Chinese BeiDou Navigation Satellite System is
proven to be of comparable accuracy to the Global Positioning System (GPS), which
is at centimeter level for the horizontal components and sub-decimeter level for the
vertical component. But the BeiDou precise point positioning (PPP) shows its
limitation in requiring a relatively long convergence time. In this study, we develop
a numerical weather model (NWM) augmented PPP processing algorithm to
improve BeiDou precise positioning. Tropospheric delay parameters, i.e., zenith
delays, mapping functions, and horizontal delay gradients, derived from shortrange forecasts from the Global Forecast System (GFS) of the National Centers for
Environmental Prediction (NCEP) are applied into BeiDou real-time PPP. Observational data from stations that are capable of tracking the BeiDou constellation from
the IGS Multi-GNSS Experiments (MGEX) network are processed, with the introduced NWM augmented PPP and the standard PPP processing. The accuracy of
tropospheric delays derived from NCEP is assessed against with the IGS final
tropospheric delay products, Fig. 3.73. The positioning results show that an
improvement of convergence time up to 60.0% and 66.7% for the east and vertical
components, respectively, can be achieved with the NWM augmented PPP solution
compared to the standard PPP solutions, while only slight improvement of the
solution convergence can be found for the north component Fig. 3.74. A positioning
accuracy of 5.7 cm and 5.9 cm for the east component is achieved with the standard
PPP that estimates gradients and the one that estimates no gradients, respectively, in
comparison to 3.5 cm of the NWM augmented PPP, showing an improvement of
38.6% and 40.1%. Compared to the accuracy of 3.7 cm and 4.1 cm for the north
component derived from the two standard PPP solutions, the one of the NWM
augmented PPP solution is improved to 2.0 cm, by about 45.9% and 51.2%. The
Fig. 3.73 (a) The time series of NCEP and IGS ZTD at station BRST for September, 2015 – day of
year 244–272. The NCEP ZTD are shown in red, and IGS ZTD in blue. (b) Distribution of ZTD
differences between NCEP and IGS
3 Advanced GNSS Processing Techniques (Working Group 1)
145
Fig. 3.74 The BeiDou RT-PPP solutions at station GMSD (Japan, 30.56 N, 131.02 E) on
September 1, 2015 (DOY 244 of 2015). The NWM augmented PPP solution is shown in red, the
standard PPP solution that estimates gradients in blue, and the standard PPP solution that estimates
no gradients in green
positioning accuracy for the up component improves from 11.4 cm and 13.2 cm with
the two standard PPP solutions to 8.0 cm with the NWM augmented PPP solution, an
improvement of 29.8% and 39.4%, respectively.
3.5.8
Using External Tropospheric Corrections to Improve
GNSS Positioning of Hot-Air Balloon16
P. Václavovic
Geodetic Observatory Pecný, Research Institute of Geodesy, Topography and
Cartography, Zdiby, Czech Republic
e-mail: [email protected]
16
Parts from this section were previously published in Vaclavovic et al. (2017).
146
J. Douša et al.
J. Douša
Geodetic Observatory Pecný, RIGTC, Ondřejov, Czech Republic
e-mail: [email protected]
M. Eliaš
Geodetic Observatory Pecný, Research Institute of Geodesy, Topography and
Cartography, Zdiby, Czech Republic
e-mail: [email protected]
Tropospheric delay of a GNSS signal has to be considered in any precise positioning
solution. When a receiver remain in a static position, which is the case of all
permanent stations, coordinates are significantly constrained, and the correlation
between station height and ZTD is not critical. Moreover, tropospheric parameters
can be also constrained during stable weather condition. However, when a receiver is
in movement, particularly in vertical direction, the height-ZTD mutual correlation
can significantly decrease an accuracy of position results. It is particularly true for
Precise Point Positioning because any biases are neither eliminated nor reduced. It
should be noted, that known precise troposphere can improve also static receiver
positioning, namely in terms of reducing convergence time due to diminishing state
vector dimension and better observation model after cold start. Consequently,
introducing ZTD from an external source can play important role in making the
solution more robust and accurate. Sources for the tropospheric parameters can be
models produced on the bases of meteorological data, tropospheric products from
GNSS analyses, or their optimal combination.
Geodetic Observatory Pecny has demonstrated how the correlation between
height and ZTD critically influences position estimates of kinematic receiver. For
this purpose, an experiment with a hot air balloon carrying a GNSS receiver together
with several meteorological sensors has been arranged (Fig. 3.75). The receiver with
other necessary equipment were mounted on the balloon basket and flied more than
2000 meters above the earth surface (Václavovic et al. 2017). Figure 3.76 demonstrates a vertical profile of the flight divided into five flight phases: (1) Initial,
(2) Ascent, (3) Descent, (4) Landing, and (5) Finish. In order to obtain stable solution
without initial convergence typical for the PPP method, the receiver remained in its
static position during the initial and finish flight phases. Collected multi-GNSS
(GPS, GLONASS, Galileo, BeiDou) high rate observations with meteorological
measurements (atmospheric pressure, relative humidity, air temperature) were
processed by the G-Nut/Geb software using the PPP method, which requires modeling all errors with the highest accuracy. This is accomplished using corrections from
external precise models or products, such as satellite orbits and clocks, satellite
attitude models, atmospheric delays, receiver and satellite antenna phase centre
offsets and variations, relativistic and phase windup effects. Experimental campaign
observations were processed utilizing real-time orbits and clocks of GPS and
GLONASS satellites provided by the Centre National d’Etudes Spatiales, CNES
(Laurichesse et al. 2013). The products are disseminated in a real-time stream, and
3 Advanced GNSS Processing Techniques (Working Group 1)
147
Fig. 3.75 Mounting GNSS antenna and meteorological sensors on the balloon basket (left) and the
balloon during ascending (right)
Fig. 3.76 Experimental
vertical profile and
definition of flight phases
they had been stored in daily files using the BKG NTRIP Client (BNC) (Weber et al.
2016).
We applied two principal strategies for ZTD modeling: (1) estimating along with
all other unknowns in the adjustment, (2) introducing from an external tropospheric
model. The tropospheric model used in this study was the Weather Research and
Forecasting (WRF) model provided by the Institute of Computer Science, Academy
of Science of the Czech Republic. A methodology for deriving ZTD from such WRF
model has been developed at GOP and published in Douša and Eliaš 2014.
Estimating ZTD simultaneously with the rover height resulted in a strong dependency of kinematic solution on the ZTD random walk noise setting. The effect was
attributed to a strong mutual correlation described by the correlation coefficient
reaching up to 0.9 for a very loose ZTD constraining spectral density 500 mm/
h1/2 of the stochastic process random walk (red line in Fig. 3.77). Depending on ZTD
random walk setting, the rover height discrepancies reached up to 30–50 cm. The
study suggests that ZTD should be tightly constrained in a vertically kinematic
solution to stabilize the results; however the constraining influences the accuracy
of estimated height in case of strong rover dynamics. The solution using not only
GPS satellites but also GLONASS became more stable due to a better satellite
148
J. Douša et al.
Fig. 3.77 Dependence of
the correlation coefficient
between ZTD and height on
random walk settings
geometry and a better decorrelation of estimated parameters. To achieve the best
accuracy, a careful offline analysis of an optimal random walk setting for estimated
ZTD should be done and, optionally, dynamically adapted. Findings of such analyses are not available in real-time, however, when a precise external tropospheric
model is applied, the offline analyses is not necessary.
The precise tropospheric model improved mainly solutions under the conditions
of poor satellite geometry which was simulated by reducing available GPS constellation. Therefore, a significantly smaller dependence of positioning precision on the
ZTD constraining was observed when using multi-constellation observation compared to single-constellation. Kinematic processing is usually difficult in real-time
because receiver movement cannot be sufficiently predicted; therefore, large noise
has to be introduced in the Kalman filter prediction. However, when post-processing
is applicable the backward smoothing can be applied (Václavovic and Douša 2015).
It the experiment, such improvement was described by the factor of two. The
combination of multi-GNSS observations and the backward smoothing algorithm
reached the best agreement between different solution variants using different
troposphere constraining.
As an optimal setting for ZTD constraining is not able to recommend, precise
kinematic positioning may benefit from external tropospheric corrections. Current
accuracy of NWM forecasts already provides corrections at the centimeter level,
which is similar or even better when compared to values estimated from GNSS data,
in particular when a standalone GNSS is used. The NWM-driven PPP solution of our
vertical experiment resulted in 9–12 cm and 5–6 cm uncertainties in the rover
altitude using the Kalman filter and the backward smoothing, respectively. Compared to the standard PPP, it indicates better performance by a factor of 1–2
depending on the availability of GNSS constellations, the troposphere constraining
and the processing strategy used. To conclude, if observation conditions are difficult,
such as with high GDOP values, external corrections from the augmented tropospheric model can significantly improve the robustness and the accuracy of the rover
height in precise positioning of prevailing vertical dynamics. More details can be
found in Václavovic et al. 2017.
3 Advanced GNSS Processing Techniques (Working Group 1)
3.5.9
149
Real-Time PPP Augmented with High-Resolution
NWM Model Data17
K. Wilgan
Wrocław University of Environmental and Life Sciences, Wrocław, Poland
e-mail: [email protected]
T. Hadaś
Wrocław University of Environmental and Life Sciences, Wrocław, Poland
e-mail: [email protected]
P. Hordyniec
Wrocław University of Environmental and Life Sciences, Wrocław, Poland
e-mail: [email protected]
J. Bosy
Wrocław University of Environmental and Life Sciences, Wrocław, Poland
e-mail: [email protected]
We proposed a high-resolution model of troposphere corrections for Poland based
on numerical weather prediction (NWP) model and GNSS data from postprocessing. The chosen NWP model is Weather Research and Forecasting (WRF).
In current configuration, the model outputs are given in a form of a dense horizontal
grid (4 km 4 km) with 47 height levels. Another factor that can impact the
positioning accuracy and convergence time is the choice of mapping functions
used to reduce the zenith delay to the slant delay. The most commonly used VMF
are based on the ECMWF model with spatial resolution of about 40 km and temporal
resolution of 6 h. In this article, the mapping functions based on a WRF model with
higher spatial and temporal resolutions were proposed.
The tropospheric corrections model was applied into real-time PPP software
GNSS-WARP (Wroclaw Algorithms for Real-time Positioning) developed at Institute of Geodesy and Geoinformatics, Wroclaw University of Environmental and Life
Sciences. The study was conducted for 14 Polish EPN stations during three periods,
with different troposphere conditions.
The performance tests were conducted in six GNSS data processing variants,
including two commonly used variants using a priori ZTD and mapping functions
from UNB3m and VMF1-FC models, one with a priori ZTD and mapping functions
calculated directly from WRF model and three variants using the aforementioned
mapping functions but with ZTD model based on GNSS and WRF data used as a
priori troposphere and to constrain tropospheric estimates. The application of a highresolution WRF/GNSS-based ZTD model and mapping functions results in the best
agreement with the official EPN coordinates. Three types of coordinates: static,
kinematic and reinitialized kinematic were estimated. In both, static and kinematic
17
Parts from this section were previously published in Wilgan et al. (2017b).
150
J. Douša et al.
Fig. 3.78 Mean 3D biases and 3D StDev of static coordinate residuals (estimated – EPN official)
averaged from 14 Polish EPN stations
mode the application of high-resolution WRF/GNSS-based model resulted in an
average reduction of 3D bias by 20 and 10 mm respectively, but an increase of 3D
standard deviations by 1.5 and 4 mm respectively (Figs. 3.78 and 3.79). This
approach also shortens the convergence time, e.g. for a 10 cm convergence level,
by 13% for the horizontal components and by 20% for the vertical component
(Fig. 3.80). More details can be found in Wilgan et al. 2017a, b.
3.5.10 Validation and Implementation of Direct Tropospheric
Delay Estimation for Precise Real-Time Positioning
L. Yang
University of Nottingham, Nottingham, UK
e-mail: [email protected]
3 Advanced GNSS Processing Techniques (Working Group 1)
151
Fig. 3.79 Mean 3D biases and 3D StDev of kinematic coordinate residuals (estimated – EPN
official) averaged from 14 Polish EPN stations
C. Hill
University of Nottingham, Nottingham, UK
e-mail: [email protected]
J. Jones
Met Office, Exeter, UK
e-mail: jonathan.jones@metoffice.gov.uk
Unmitigated tropospheric delay remains one of the major error sources in PPP. Due
to the lack of real information, conventional empirical models (for both ZTD and
mapping functions) have limitations in terms of positioning accuracy and convergence time, especially during active tropospheric conditions. Aiming to overcome
these limitations and improve PPP performance, we investigated the feasibility of
integrating the real atmospheric condition from NWM into PPP, as an external
152
J. Douša et al.
Fig. 3.80 Convergence time for different levels of convergence for horizontal (up) and vertical
(bottom) components and a percentage of converged solutions
correction service. The main challenge is to capture and abstract the most effective
information from NWP data, and transmit to the user in a bandwidth-economic way.
The fundamental concept proposed is a multiple-tiered data transmission structure. The tiers will provide quality information in different levels of details. The
content of the tiers is shown in Fig. 3.81. In the whole service area, the atmospheric
conditions are monitored in grid cells. The higher tiers are only produced and
transmitted where/when they could bring additional accuracy compared to the
lower tiers. An indication type of flag map is produced in real-time to inform users
of the number of active tiers available at the users’ grid cell. This tiered transmission
scheme could effectively reduce the transmission bandwidth consumed, while keeping a certain guarantee of accuracy.
In this project, the state of the art Unified Model (UM) produced by the UK Met
Office (UKMO) is selected for the underpinning NWP. To investigate the
3 Advanced GNSS Processing Techniques (Working Group 1)
153
Fig. 3.81 Tiered
troposphere delay correction
structure
performance of the proposed correction service, statistical analysis is carried out
using data over 12 months in 2014 with a 6-h interval at 108 CORS stations in
the UK.
3.5.10.1
Impact of the Multiple Tiers
In Fig. 3.82, the RMS errors of different ZTD estimation methods are compared,
(using DD solution as the reference). Three different empirical ZTD estimations
(Tier 0, which does require any external information), i.e. the MOPS model, the
GPT2 model and the ESA Blind (ESA-B) mode model, show very similar performance. While clear improvements can be observed for the higher tier solutions.
Comparing to the ESA-B solution, there is a 27.3% improvement for the ESA site
mode (Tier 1, which provides surface meteorological parameters), 45.2% for the
ESA augmented mode (Tier 2, which provides vertical lapse rates of meteorological
parameters) and 73.9% for the zenith ray tracing solution (which provides the
ultimate on-site ZTD correction). Clearly the more real information provided, the
better the ZTD estimation accuracy achieved.
Figure 3.83 provides the slant tropospheric delay (STD) error comparison
between the solutions of Tier 0 (ESA blind mode), Tier 1 (ESA site mode), Tier
2 (ESA augmented mode), Tier 3 (ZTD and mapping function correction) and Tier
4 (gradient correction). The Tier 3 solutions are further split into the ZTD only
correction and the ZTD plus mapping function correction. Ray tracing is used as the
reference. The mean error and standard deviation of each solution are plotted in
individual figures in the left, and the RMS errors of all solutions are plotted together
against elevation in the right.
It can be seen that, from the blind mode to the site mode and then the augmented
mode, both mean error and standard deviation are improved in general, which results
in the improved RMS errors in turn. The improvement from blind mode to site mode
154
J. Douša et al.
Fig. 3.82 RMS of ZTD error from different estimation methods
Fig. 3.83 TSD RMS error against elevation, for different tiers
is larger than the improvement from site mode to augmented mode indicating that,
although the real vertical rates of change are beneficial, the real surface parameters
play a more important role in the STD estimation. The ZTD correction could
effectively reduce the mean error offset at higher elevations, as well as the uncertainty in the lower elevation area. This is because if the ZTD is not estimated
properly, the error could be magnified by the mapping function. Even with the
ZTD correction, the RMS error at 10 elevation is still higher than 10 cm. Both
mapping function and gradient correction could reduce the error in the lower
elevation, while the former has a stronger impact. Therefore, it is recommended
that the mapping function shall be corrected earlier than the gradient. It can also be
seen that at the 10 elevation, the improvement from the mapping function correction
3 Advanced GNSS Processing Techniques (Working Group 1)
155
will be larger than the improvement from the ZTD correction, and could reduce the
RMS error to centimetre level.
3.5.10.2
Flag Map
One of the key features in our proposed tiered troposphere delay correction scheme
is the flag map. This map is updated in real time, and provides users with local
information, that to which correction tier the user should use to guarantee a certain
STD estimation accuracy. The size of this flag map is quite small, thus is suitable for
broadcast in a transmission medium with limited bandwidth, such as a communication satellite. Figure 3.84 shows an example of the tiered structure flag map, which is
consist of 60 by 90 grid squares (~17 km 17 km) for the UK region.
Figure 3.85 shows the total occurrence probability of each tier, from the statistic
of 12 months in the UK CORS stations. Each subfigure presents different elevation
settings, ranging from 10 to 70 . In each subfigure, the x axis is various thresholds
that could be pre-set as the accuracy targets, and the y axis gives the chance
(in percentage) of each correction tier being required.
At high elevations, a simple empirical estimation approach would be able to meet
a lenient accuracy target. With an increasingly strict threshold, real information is
required to replace empirical values to suppress the modelling error. When the
elevation gets lower, the STD modelling assumptions based on the empirical
mapping function and balanced local atmosphere profile will gradually become
less valid, and the error they bring cannot be ignored.
Fig. 3.84 Example of the flag map for the tiered structure (06th Oct 2014 0600 GMT)
156
J. Douša et al.
Fig. 3.85 The chance of each tier to be applied, against different threshold
3.5.10.3
PPP Improvement
Figure 3.86 uses four storm event cases (at each row) to indicate the PPP performance improvement from the proposed correction scheme, in which the red lines
indicate the traditional PPP solution using empirical tropospheric modelling and the
green lines indicate PPP with external tropospheric correction. In the left column, the
convergence times for 10 cm (dash line) and 5 cm (solid line) are shown. And in the
right column, the positioning standard deviation in 3D (solid line) and height
component (dash line), at the end epoch after the 1-h PPP solution, are shown.
The positive impact of the proposed external correction can be observed from both
aspects.
3.5.10.4
Bandwidth
Bandwidth is a key issue in for the proposed troposphere correction scheme, since if
the bandwidth required is too high to be transmitted, the whole scheme is not
practical. A customized transmission has been designed in this project, which only
transmits the necessary information to where it is required and hence effectively
reduce the total required bandwidth and also help in multiplexing the transmission
channel.
3 Advanced GNSS Processing Techniques (Working Group 1)
157
Fig. 3.86 Comparison of PPP convergence time and standard deviation after 1 h
Table 3.21 Summary of transmission design details for each tier, for UK coverage
Tier
Bandwidth (kbps)
1
4.74
2
2.60
3
0~5.21
4
0~2.60
5
397~531
In Table 3.21 the required transmission bandwidth is estimated. It can be seen
that, for the UK coverage, the total required bandwidth for tier 1–4 will be
7.34~15.15 kbps, and 397~529 kbps for Tier 5. Meanwhile, the size of each flag
map message will be 110 Bytes, which could be easily broadcasted. Tier 1–4 and the
flag map message could be easily fit into communication satellite bandwidth nowadays. Tier 5 information is better to be transmitted via 4/5G network and through a
two-way communication once required.
3.5.10.5
Summary
• Real time ray tracing could provide an accurate STD estimation, and thus
improve PPP results.
• Although the required troposphere condition data for the real time 3D full ray
tracing is huge, a tiered correction structure and flag maps mechanism has been
158
J. Douša et al.
designed to reduce data transmission volume effectively. The data are only
transmitted for the time and place when they are required.
• Statistical results show that only very occasionally is the full ray tracing correction required, and for most of the time, the mapping function correction could
allow the user to satisfy STD estimation accuracy.
• The total bandwidth required for the UK coverage is 7.34~15.15 kbps for Tier
1–4, and 397~529 kbps for Tier 5. The flag map message size is 110B each. Apart
from Tier 5, other information could be easily transmitted via communication
satellite.
• The current drawback of the whole proposed solution is the NWP data resolution,
especially in time (currently 6-h interval).
3.6
GNSS Data Reprocessing for Climate
One of the important tasks of the WG1 of the COST Action was the homogeneous
reprocessing of GNSS data for climate applications in support of WG3. Main focus
of WG1 was to study the impact of precise models and processing strategy for
providing a long-term tropospheric parameters with an optimal accuracy, a high
reliability and a consistent quality. This section gives an overview of the
reprocessing activities during the COST Action period, while WG3-related activities
are described in Chap. 5, including reprocessing product validation, combination,
quality control and homogenization.
This section introduces contributions to the GNSS second reprocessing
performed within the IAG sub-commission for the European Reference Frames
(EUREF) and within the IGS TIGA campaign. Although these were primarily
designed for contribution to geodetic applications, within the COST Action a
significant effort was additionally dedicated to an optimal estimation of tropospheric
parameters on regional and global scales. A special activity supporting directly the
climate community has been then started by establishing the GRUAN Central
Processing Centre at GFZ for a long-term ground-based GNSS precipitable water
estimation.
3.6.1
EUREF Repro2 Contribution of Swisstopo
E. Brockmann
Swiss Federal Office of Topography swisstopo, Wabern, Switzerland
e-mail: [email protected]
3 Advanced GNSS Processing Techniques (Working Group 1)
159
Fig. 3.87 Key parameters of swisstopo repro2
The Swisstopo finished a reprocessing in 2014 covering a homogeneous processing
of a time span starting beginning 1996 till almost the end of 2014. Some key
parameters are given in Fig. 3.87. The number of stations increased from 20 to
170. The stations are mostly located in central Europe. Some boundary stations were
included to enable a better decorrelation between troposphere parameters with
station height. The number of satellites increased from 25 in 1996 to 65 in 2014.
Since 2004, the number of satellites increased due to the improved GLONASS
constellation.
The reprocessing was performed with the BSW52 using most up-to-date models
in a homogeneous way. Orbit and earth rotation parameters were used from the
CODE repro2 products. The important processing options are given in Table 3.22.
Figure 3.88 shows that daily solutions are even more stable in view of the Helmert
transformation parameters than the previous weekly solutions.
Overlapping 3-day solutions are calculated to optimize the ZTD estimates at
midnight (see Fig. 3.89).
Moreover, variations on the processing options were carried out in order to find
the best possible modelling options and in order to do some sensitivity studies:
• GMF/GPT mapping, atmosphere loading (ATL) + IGS08 group antenna PCV
(submitted as LP0 to EUREF: Oct. 23. 2014)
• VMF mapping, IGS08 antenna model but with individual antenna PCV where
available, atmospheric tidal loading ATL, non-tidal atmospheric pressure loading
ALP (submitted to EUREF as LP1: March 20, 2015).
• Same as above, but GPS-only (idea borne on Thessaloniki workshop 2014,
results generated in June 2015 and presented at Wroclaw COST meeting end of
September 2015)
160
J. Douša et al.
Table 3.22 Basic processing options used for swisstopo repro2
Software
Satellite systems
Elevation cutoff
angle
Observation
weighting
Antenna
Troposphere
Troposphere
gradients
Tides
Conventions
Ocean tides
Gravity field
Ionosphere
Reference frame
Network
Time span
Orbits/EOP
BSW52 (+)
GPS + GLO (ab 2004)
3
COSZ elevation-dependent weighting
I08 absolute antenna model (group values)
GMF and DRY GMF mapping for the a priori values and while estimating
hourly ZPD parameters using WET GMF
Chen Herring for tropospheric gradient estimation
Atmospheric tidal loading applied
IERS2010
FES2004
EGM08
CODE 2-h resolution; including higher order terms
IGb08
Max. 180 stations
DOY 007, 1996 till DOY 207, 2014
CODE reprocessing series 2011 (till DOY 106, 2011) and CODE
reprocessing series 2013 (till DOY 362, 2013), CODE operational series in
2014
• The impact of additional GLONASS observations on the long-term (comparison
of the before mentioned last two repro2 series) was especially analyzed in this
project. The impact of the additional GLONASS observations is negligible when
analyzing ZTD trends. Further conclusions can be drawn from the comparisons
between different ZTD time series:
• Formal errors of ZTD estimates are smaller with more GLONASS observations
(max. 10%) included
• Influence of GMF versus VMF: no significant rate, standard deviation
0.5–2.5 mm ZTD (109 sites with long time series)
• Difference to CODE global Repro: no significant rate, standard deviation
2.0–5.0 mm ZTD (39 sites with long time series);
• Influence additional GLONASS observations: no significant rate, standard deviation: 0.4–1.5 mm (111 sites) (MODA exception). This is In fact a little higher
(also GPS-only time counted), but strongly dependant on the used a priori ZTD
constraints, there is only a statistical effect but no significant bias.
Comparisons with e.g. radio sondes have much bigger differences (see Fig. 3.90).
In case of PAYE we see a standard deviation of 8–9 mm ZTD. The estimated rate of
5 mm/10 years. is also dependent on changes at the radio sonde station. In 2009, a
3 Advanced GNSS Processing Techniques (Working Group 1)
161
Fig. 3.88 Stability of the solutions expressed in Helmert Parameters between each individual
solution with respect to the combined solution (operational older weekly solution top, swisstopo
repro2 daily solution bottom)
new humidity sensor was used. Trend estimates on the raw ZTD time series are quite
sensitive – especially the various antenna changes at the GNSS stations generate
quite significant jumps in station heights as well as in the ZTDs.
162
J. Douša et al.
Fig. 3.89 3-day solutions to optimize the ZTD estimates at midnight
Zenith Total Delay Difference for station PAYE in [mm]
PAYE: R2H3 minus RS, N=7377
linear fit, std=8.3 mm
40
20
0
–20
Repro2, VMF, 3D: 8.3 mm std
–40
2004
2006
2008
Radiosonde: new
humidity sensor
31/03/2015 15:00
40
2010
YEAR
2012
2014
2016
PAYE: R2G1 minus RS, N=7377
linear fit, std=8.9 mm
20
0
–20
Repro2, GMF, 1D: 8.9 mm std
–40
2004
2006
2008
2010
YEAR
2012
2014
2016
31/03/2015 14:50
Fig. 3.90 Swisstopo Repro2 (VMF 3 days upper diagram, GMF 1 day lower diagram) compared
with radio sonde derived ZTD estimates for station PAYE
3 Advanced GNSS Processing Techniques (Working Group 1)
3.6.2
163
EUREF Repro2 Assessment of GOP Processing
Variants18
J. Douša
Geodetic Observatory Pecný, RIGTC, Ondřejov, Czech Republic
e-mail: [email protected]
P. Václavovic
Geodetic Observatory Pecný, Research Institute of Geodesy, Topography and
Cartography, Zdiby, Czech Republic
e-mail: [email protected]
The results of the 2nd reprocessing of all data available from 1996 to 2014 from all
stations (>300) of the European GNSS permanent network are introduced as
performed at the Geodetic Observatory Pecný (GOP) (Douša et al. 2017). The
reanalysis was completed during the 2nd EUREF reprocessing to support mainly
the realization of a new European terrestrial reference system. The BSW52 (Dach
et al. 2015) was used together with CODE precise products (Dach et al. 2014).
Within the COST ES1206 project, a new set of GNSS tropospheric parameter
time series was provided for applications in climate research. To achieve this goal,
we improved our strategy for combining tropospheric parameters over three consecutive days that guarantees a continuity of all estimated tropospheric parameters,
zenith tropospheric delays (ZTD) and tropospheric horizontal linear gradients, at all
mid-nights and during transitions of GPS weeks. Basic characteristics of the 2nd
reprocessing are provided in Table 3.23.
Within the reprocessing, we performed additionally seven solution variants in
order to study an optimal troposphere modelling, Table 3.24. We assessed variants in
terms of the coordinate repeatability by using internal evalautions of coordinate
repatability. Then, we compared ZTDs and tropospheric horizontal gradients with
independent values obtained from ERA-Interim numerical weather reanalysis (Dee
et al. 2011).
Generally, the results of the GOP Repro2 yielded improvements of approximately
50% and 25% for the horizontal and vertical component repeatability, respectively,
when compared to the results of the GOP Repro1 solution. Vertical repeatability was
reduced from 4.14 mm to 3.73 mm when using the VMF1 mapping function (Boehm
et al. 2006a, b), a priori ZHD, and non-tidal atmospheric loading corrections from
actual weather data. Increasing the elevation cut-off angle from 3 to 7 /10
increased RMS errors of residuals from the coordinates’ repeatability. These findings
were confirmed also by the independent assessment of tropospheric parameters
using NWM reanalysis data. The differences of particular solutions were statistically
analysed to demonstrate the impact of the different modelling on ZTDs, see
18
Parts from this section were previously published in Douša et al. (2017).
164
J. Douša et al.
Table 3.23 Characteristics of GOP reprocessing models
Analysis
options
Products
Observations
Reference
frame
Antenna
model
Troposphere
Ionosphere
Loading
effects
Description
CODE precise orbit and earth rotation parameters from the IGS 2nd reprocessing.
Dual-frequency code and phase GPS observations from L1 and L2 carriers.
Elevation cut-off angle 3 , elevation-dependent weighting 1/cos2 (zenith),
double-difference observations and with 3-min sampling rate.
IGb08 realization, core stations set as fiducial after a consistency checking.
Coordinates estimated using a minimum constraint.
GOP: IGS08_1832 model (receiver and satellite phase Centre offsets and
variations).
A priori zenith hydrostatic delay/mapping function: GPT/GMFh (GO0) and
VMF1/VMF1 h (GO1-GO6). Estimated ZWD corrections every hour using
VMF1w mapping function; 5 m and 1 m for absolute and relative constraints,
respectively. Estimated horizontal NS and EW tropospheric gradients every 6 h
(GO0-GO5) or 24 h (GO6) without a priori tropospheric gradients and
constraints.
Eliminated using ionosphere-free linear combination (GO0-GO6). Applying
higher-order effects estimated using CODE global ionosphere product (GO5).
Atmospheric tidal loading and hydrology loading not applied. Ocean tidal
loading FES2004 used. Non-tidal atmospheric loading introduced in advanced
variants from the model from TU-Vienna (GO4-GO6).
Table 3.24 GOP solution variants for the assessment of selected models and settings
Solution
ID
GO0
GO1
GO2
GO3
GO4
GO5
GO6
Specific settings and
differences
GMF and 3 cut-off
VMF1 and 3 cut-off
¼GO1; 7 cut-off
¼GO1; 10 cut-off
¼GO1; atmospheric loading
¼GO4; higher-order
ionosphere
¼GO4; 24-h gradients
Remarks and rationales
Legacy solution for Repro1
New candidate for Repro2
Impact of elevation cut-off angle
Impact of elevation cut-off angle
Non-tidal atmospheric loading applied
Higher-order ionosphere effect not applied
Stacking tropospheric gradients to 24-h
sampling
Table 3.25. Although the VMF1 mapping function outperformed GMF in term of
standard deviation for the elevation cut-off angle of 3 (GO0 vs GO1), the impact is
smaller than for the others effects, i.e. due to elevation cut-off angle (GO2 vs GO1,
GO3 vs GO1) and non-tidal atmospheric modelling (GO4 vs GO1). The results get
worse for both ZTD and tropospheric gradients when raising the elevation cut-off
angle from 3 to 7 (GO2) or 10 (GO3). No significant impact of modelling of highorder ionospheric effects (GO5) was observed; the effect is systematic for regional
network and was eliminated by applying fiducial stations. Figure 3.91 shows a
significant improvement for gradients (negligible impact on ZTD) when combining
3 Advanced GNSS Processing Techniques (Working Group 1)
165
Table 3.25 Median, minimum and maximum values of total ZTD biases and standard deviation
over all stations
Compared
solution
variants
GO1 vs GO0
GO2 vs GO1
GO3 vs GO1
GO4 vs GO1
GO5 vs GO4
GO6 vs GO4
ZTD bias
median
[mm]
0.36
+0.03
+0.03
+0.05
0.02
0.02
ZTD bias
min [mm]
1.52
0.81
2.22
3.29
0.31
0.23
ZTD bias
max [mm]
+0.70
+1.66
+2.66
+5.55
+0.07
+0.16
ZTD sdev
median
[mm]
2.01
0.66
1.10
1.37
0.07
1.24
ZTD sdev
min [mm]
0.69
0.15
0.31
0.68
0.04
0.76
ZTD sdev
max [mm]
3.82
1.29
2.04
4.72
0.30
2.46
Fig. 3.91 Monthly means of bias and standard deviation of tropospheric horizontal north (N-GRD)
and east (E-GRD) gradients compared to those obtained by ERA-Interim. Error bars indicate
standard errors of mean values over all compared stations plotted from the zero y-axis to emphasise
seasonal variations and trends. Error bars are displayed for north gradients only, however, being
representative for the east gradients too
gradient parameters from the original 6-h into 24-h time resolution without applying
constrains. The 6-h gradients were able to absorb some remaining errors in GNSS
model.
We also studied a temporal and spatial variation of ZTD differences from seven
solution variants, for details see Douša et al. 2017. Figure 3.92 shows latitudinal
166
J. Douša et al.
Fig. 3.92 Dependence of ZTD biases (blue) and standard deviations (red) from inter-comparisons
of GOP 2nd reprocessing solution variants on station latitude. Note different y-range for the
GO5 vs. GO4 comparison
dependence of the ZTD differences, while dependences on height, time and geographical location are not showed here.
Based on the reanalysis variants, we recommend using low-elevation observations together with the VMF1 mapping function, precise a priori ZHD values and
consistent model of non-tidal atmospheric loading. The daily piecewise linear
function model showed better stability for estimated gradients while did not indicate
a worse repeatability of coordinates estimates. For saving the time, we could
recommend such (unconstrained) approach for an optimal modelling of the firstorder tropospheric asymmetry. The main difficulties faced during the 2nd
reprocessing was the quality of the GNSS historical data containing a large variety
of problems. To provide high-accuracy GNSS tropospheric products, the elimination
of problematic data was more critical compared to estimating coordinates on a daily
only. The 3rd EUREF reprocessing should thus exploit more complex data quality
control in order to optimize the reprocessing and the quality of tropospheric
products.
3 Advanced GNSS Processing Techniques (Working Group 1)
3.6.3
167
CORDEX.be Reprocessing
E. Pottiaux
Royal Observatory of Belgium, Brussels, Belgium
e-mail: [email protected]
B. Van Schaeybroeck
Royal Meteorological Institute of Belgium, Uccle, Belgium
e-mail: [email protected]
The CORDEX.be stands for “COordinated Regional Climate Downscaling EXperiment and beyond” (http://cordex.meteo.be/). This national project brings together
the Belgian climate and impact modelling research groups into one research network
as the first step towards the realization of climate services. The philosophy of
CORDEX.be is inspired by the World Climate Research Programme (WCRP)
project CORDEX (“COordinated Regional Climate Downscaling Experiment”,
http://www.cordex.org/) project, but – as the “.be” in the acronym indicates – the
project aims to go beyond for Belgium.
CORDEX.be has 4 main targets (see http://cordex.meteo.be/meteo/view/en/
29026726-Targets.html): (1) to contribute to the international CORDEX project,
(2) to go beyond by running 4 High-Resolution Limited Area Models (ALARO-0,
MAR, and two flavours of COSMO-CLM, all with a resolution of 4 km, but with
different forcing strategies), (3) to go beyond by running 4 Local-Impact Models
(Wave and storm surge, urban, crop Isoprene or vegetation emission models), and
(4) to infer the climate uncertainties to the Belgian level. In addition, a specific task is
dedicated to the validation of the high-resolution climate simulations using GNSSderived products. The main objective of this task is to go beyond the standard
verification procedure of climate simulations. The traditional manner is to compare
the results from climate runs with long-term ground-based surface meteorological
observations. Instead, here we will implement a verification based on products
estimated from continuously operating GNSS (Global Navigation Satellite Systems,
such as the American GPS) stations.
Therefore, a careful reprocessing of the historical observations at about
320 world-wide GNSS stations (Fig. 3.93, left) has been carried out with a focus
on the period 2000–2010 (i.e. the period requested by the CORDEX.be partners for
the assessment of the climate models), allowing both the validation of the standard
CORDEX runs and the high-resolution runs. Unfortunately, from these 320, only
20 GNSS stations (Fig. 3.93, right) which were operating almost continuously and
providing high-quality observations throughout the requested validation period
(2000–2010) are located in the restricted high-resolution climate model domain,
but many other European GNSS sites can still be use to validate the standard 12.5
and 50 km resolution European CORDEX runs.
First results of this GNSS-based validation of the hourly ZTDs derived from the
high-resolution model output is given in Fig. 3.94 (only for two climate models,
MAR and COSMO-CLM, but with different forcing scenarios for MAR: NCEP,
168
J. Douša et al.
Fig. 3.93 Left: All GNSS stations included in the reprocessing. Right: GNSS stations located in the
high-resolution climate domains of CORDEX.be
COSMO-CLM KUL (ERA-I)
MAR (ERA-I forcing)
Model / observed variability ZTD
Anomaly correlation of hourly ZTD
MAR (NCEP forcing)
MAR (ERA20C forcing)
90%
80%
70%
60%
50%
40%
30%
20%
Feb
Apr
Jun
MONTH
Aug
Oct
Dec
110%
105%
100%
95%
5
10
15
20
Hour of the day
Fig. 3.94 Anomaly correlation (left) and variability ratio (right) of hourly ZTD values between
different model simulations and the GNSS-derived observations for the different months of the
year. The models considered are the H-Res simulations of COSMO-CLM (driven by ERA-Interim),
and MAR (driven by NCEP-NCAR-v1, ERA-Interim and ERA20C). Averages are taken over seven
Belgian GNSS stations
ERA20C and ERA-Interim). It is seen that all model-based ZTDs correlate pretty
well with the GNSS-based ZTDs. The annual cycle (Fig. 3.94, left) from all models
shows a better agreement during the winter months while more pronounced departure is visible during the summer months. Very similar yearly patterns in the
anomaly correlation is visible for the two climate models and the various forcing
scenarios represented. For the daily cycle (Fig. 3.94, right), also very similar daily
patterns in the “model over observed variability of the ZTD” is visible in all cases,
with a peak drop in the mid-late afternoon. The current working hypothesis is that
the higher the water vapour variability, the higher the departure between model and
observation is visible. In terms of forcing the climate model runs: as expected for the
3 Advanced GNSS Processing Techniques (Working Group 1)
169
MAR results, results with ERA-Interim forcing are better than forcing with NCEPNCAR-v1 reanalysis which in turn are better ERA20C forcing. The better scores for
MAR with respect to COSMO-CLM may be caused by contrasting coupling strategies. More specifically, whereas MAR is directly forced at the boundaries by
reanalysis, COSMO-CLM uses an additional nesting to obtain the highest spatial
resolution of 2.8 km. This validation work is still ongoing and more detailed
assessments will be undertaken in the future.
3.6.4
GRUAN Reprocessing
G. Dick
GFZ German Research Centre for Geosciences, Helmholtz Centre Potsdam,
Potsdam, Germany
e-mail: [email protected]
F. Alshawaf
GFZ German Research Centre for Geosciences, Potsdam, Germany
e-mail: [email protected]
The Global Climate Observing System (GCOS) Reference Upper Air Network
(GRUAN) of the World Meteorological Organization (WMO) is an international
reference observing network, designed to meet requirements of climate research.
Upper air observations within GRUAN will provide long-term high-quality climate
records. A GNSS receiver is part of the GRUAN station equipment with highest
priority for deriving PWV. Due to its long-term experience in GNSS data
processing, GFZ was selected by WMO as a Central GRUAN GNSS Data
Processing Centre. GFZ operates also a number of GNSS stations on GRUAN
sites, see Fig. 3.95.
Fig. 3.95 GRUAN network with GNSS stations on GRUAN sites operated by GFZ (in red)
170
J. Douša et al.
Fig. 3.96 GNSS-derived PWV results from reprocessed data for 2001–2016 compared with
radiosonde measurements at GRUAN site Lindenberg (Germany)
The GRUAN GNSS Processing Centre at GFZ with EPOS software covers a fully
automated processing chain starting with collecting of raw GNSS observations and
resulting with climate relevant validated PWV products, which are available online.
PWV uncertainty estimation (Ning et al. 2016), comparisons with radiosonde
(RS) measurements (Figs. 3.96 and 3.97) as well as PWV trend estimation
(Fig. 3.98) are essential parts of the GNSS data analysis at GFZ for climatological
applications.
3.6.5
GFZ TIGA Reprocessing
Z. Deng
GFZ German Research Centre for Geosciences, Potsdam, Germany
e-mail: [email protected]
Being a modern geodetic measuring method, GNSS (Global Navigation Satellite
Systems) has reached an important role in geosciences. Within the scope of the Tide
Gauge Benchmark Monitoring Working Group (TIGA-WG) of the IGS, GFZ
analyses and reprocesses GNSS data of stations near tide gauges (Deng et al.
2014). This allows us to monitor tide gauges for vertical land deformations,
e.g. due to postglacial uplift. TIGA also contributes to the calibration of satellite
altimeters and the unification of height systems (Hunegnaw et al. 2017). In conjunction with circa 400 global IGS stations, GFZ processes data of almost 500 GNSS
stations near tide gauges between 1994 and today (Fig. 3.99). In the first TIGA
combination there are contributions from 3 international ACs. The GFZ TIGA
solution shows the best accuracy among the three submitted solutions (Hunegnaw
et al. 2017).
3 Advanced GNSS Processing Techniques (Working Group 1)
171
Fig. 3.97 GNSS-derived PWV results from reprocessed data for 2011–2016 compared seasonally
with radiosonde measurements at GRUAN site Ny-Alesund (Norway). Seasonal differences
between GNSS PWV and RS can be explained by different behaviour of RS during warmer and
colder months
Fig. 3.98 Example of PWV trend estimation from reprocessed GNSS data for 2000–2016 at
GRUAN site Lindenberg. Trend value is 0.31 mm/decade, sigma of the trend is 0.075 mm/decade
172
J. Douša et al.
-150˚
-100˚
-50˚
0˚
50˚
100˚
150˚
50˚
50˚
0˚
0˚
-50˚
-50˚
-150˚
-100˚
-50˚
0˚
50˚
100˚
150˚
Fig. 3.99 GNSS stations processed at GFZ in the framework of TIGA
Since the GFZ TIGA processing complied with the accords of the 2nd IGS
reprocessing campaign, our reprocessed solutions contributed to the determination
of the ITRF2014 (Rebischung et al. 2016). In that context the GFZ solution
contained the most stations among all of the nine solutions. Because it is the only
solution that contains all IGS stations, the IGS Analysis Centre Working Group
decided at the 2017 IGS Workshop in Paris, to routinely deliver the GFZ TIGA
solution to the IGS in order to ensure, that all IGS stations are contained in the
combined weekly IGS solutions.
GFZ TIGA products are available via FTP servers of GFZ and CDDIS. The GFZ
FTP server provides daily and weekly files for coordinates, orbits and Earth rotation
parameters (ftp://ftp.gfz-potsdam.de/pub/transfer/kg_igs/igstiga/solutions/). In addition to coordinate and orbit products, troposphere parameters are provided, which
can be used for climate studies (ftp://ftp.gfz-potsdam.de/GNSS/products/tiga_
repro2_tro/).
3.6.6
ULX TIGA Reprocessing19
F. N. Teferle
University of Luxembourg, Luxembourg, Luxembourg
e-mail: [email protected]
ULX, as one of the IGS Tide Gauge Benchmark Monitoring (TIGA) Working Group
AC has carried out a second reprocessing campaign in line with IGS. Using the latest
19
Parts from this section were previously published in Hunegnaw et al. (2015).
3 Advanced GNSS Processing Techniques (Working Group 1)
173
Fig. 3.100 120 selected global stations from the reprocessed TIGA solution. The selection is based
on the time length of the ZTD time series and their quality. We have only selected those sites having
a minimum length of 6 years
available bias models and methodology the different IGS ACs re-analyzed the full
history of GPS data collected by the global tracking network from 1995 to 2015. The
consortium of the British Isles continuous GNSS Facility (BIGF) and the University
of Luxembourg TIGA Analysis Centre (BLT) completed a new global solution using
<750 GPS stations. Figure 3.100 shows a map of 120 stations. As it can be seen, the
stations are globally distributed and the timeseries varies from 6 to 21 years in
length.
The re-processing follows a double difference network strategy using the BSW52
(Dach et al. 2015), incorporates recent bias model developments, the latest IERS
2010 conventions (Petit and Luzum 2010) and IGS recommendations. Further
details are detailed in (Hunegnaw et al. 2015). The selected station network included
all IGb08 core stations (Rebischung et al. 2012) and more or less the complete
archive of TIGA, which encompasses a large number of GPS stations at or near the
global network of tide gauges. The GPS data was re-processed using the CODE final
precise orbits and Earth orientation parameters. We employed the IGS08 satellites
and receiver antenna phase centre models and adopted an elevation cut-off angle of
3 (Dach et al. 2016). In our solution we make use of the VMF1 (Boehm et al. 2006a,
b) that allows to describe the atmosphere with the finest detail, leading to the highest
precision in the derived tropospheric parameters.
In BSW52, the ZHD is parameterized as a piece-wise function variation of the
delay using a piece-wise linear interpolation between temporal nodes. Observations
of atmospheric pressure at the GPS station offer high precision for the ZHD
estimates and minimize station height errors (Tregoning and Herring 2006). However, many of the TIGA and IGS stations do not possess integrated meteorological
sensors. Thus, ZHD in units of meters was a priori obtained reliably from surface
pressure data from the gridded output of the ECMWF NWP model and is provided
174
J. Douša et al.
by VMF1 using the modified Saastamoinen model, which assumes that the atmosphere is in hydrostatic equilibrium (Davis et al. 1985). We estimate the ZTD
parameters in an interval of 1 h with a loose constraint of 5 m. In addition, horizontal
gradients in the North-South and East-West directions are estimated in a 24-h
interval with the same 5 m loose relative constraint. In this manner more than two
decades of ZTD time series along with station positions are available from our
re-processing. Figure 3.100 shows a selection of 120 global stations for which we
have carried out the further analysis described in this study. However, as the station
positions are affected by on average two discontinuities per station per decade, the
ZTD time series need to be homogenized before being useful for further application.
The results for these 120 stations in terms of a statistical analysis of the periodic
signals and stochastic porperties of the related ZWD time series can be found in Klos
et al. (2018).
3.7
New Analysis Centres, Networks and Solutions
In this section, the new ACs are mainly presented, however, including also new
networks and strategies as well as the shared system for facilitating a collaboration
between existing and new ACs. Some of the new ACs are located in Eastern and
South-Eastern Europe (Bulgaria, Greece, Turkey, Romania), i.e. large areas where
no operational products existed so far, others have been established in countries like
Austria, Iceland and Portugal. All of them has developed or gained the expertise
thanks to the COST Action. Creating of new GNSS ACs thus fulfilled one of the
very important goals of the Action: to increase the observing network and to
facilitate the transfer of knowledge for establishment of new GNSS ACs. Now,
almost all of the new ACs contribute to E-GVAP – the EUMETNET EIG GNSS
Water Vapour Programme (http://egvap.dmi.dk).
3.7.1
Trop-NET System for Collaborative Ground-Based
GNSS Meteorology
J. Douša
Geodetic Observatory Pecný, RIGTC, Ondřejov, Czech Republic
e-mail: [email protected]
The Trop-NET system has been developed at the Geodetic Observatory Pecny
(GOP), Czech Republic, in support of the GNSS4SWEC transfer of knowledge.
The goal was to facilitate establishment of new analysis centres for near real-time
troposphere monitoring in support of numerical weather prediction within the
EUMETNET EIG GNSS Water Vapour Programme – E-GVAP (http://egvap.dmi.
3 Advanced GNSS Processing Techniques (Working Group 1)
175
dk). Three short-term scientific missions has been carried out within the transfer of
knowledge, see the STSM summary in Appendices.
The ground-based GNSS near real-time processing using a batch approach
requires following aspects: (1) hourly data provision, (2) predicted orbit products
and precise models, (3) efficient and robust procedure for fully automated operation,
and (4) a continuous monitoring and product evaluations. Since 1997, GOP has been
developing a flexible system for automated GNSS processing using the BSW and
BPE. The system also included data/product flow and supports scientific applications for estimating various parameters in a flexible update rate for different purposes/services:
•
•
•
•
•
•
•
near real-time GPS regional troposphere monitoring (EGVAP)
near real-time GPS + GLONASS regional troposphere monitoring (EGVAP)
near real-time GPS global troposphere monitoring (EGVAP)
ultra-rapid GPS and GLONASS orbit determination (IGS)
rapid (daily), final (weekly) GPS solution for reference frame (EUREF)
homogeneous re-processing of full EPN (EUREF)
daily/hourly data flow at the local data centre (EUREF)
All these applications are based on a common module library continuously
extended for a higher flexibility and robustness with a near real-time GNSS analysis
for the troposphere monitoring as the earliest application (Douša 2001a, b). Due to
initial limits in hardware and software, frequent instabilities of data flow, low quality
of 24-h predicted orbit products, the system designed had to be highly efficient, fully
self-supporting and maximally robust. During the decade, enhancements were done
particularly by extending the library for ultra-rapid orbit determination (Douša
2004a, b, 2010), GLONASS processing (Douša 2012), global NRT troposphere
solution (Douša and Bennitt 2013) and long-term coordinate and troposphere
re-analysis (Douša et al. 2017).
The above mentioned experience leaded to the idea of sharing the library for a
collaborative use within the GNSS4SWEC project. For this purpose, the Trop-NET
package has been completed for easier dissemination, configuration and maintenance in different environments. Main goals within the COST ES1206 were:
(a) facilitate the establishment of new analysis centres, (b) improve the product
coverage and its homogeneity over Europe, (c) give a possibility to share future
developments, and (d) enable coordinated solution updates. Currently, the TropNET pack is maintained by GOP using the Subversion repository. The system
consists of several modules supporting individual settings for different user scenarios (Fig. 3.101). The distributed processing is supported by three core modules:
(1) for data and product download and mirroring, (2) for GNSS data processing, and
(3) for product uploading. Additionally, the system includes central components
currently maintained by GOP such as software distribution and update system, NRT
product monitoring system, long-term product evaluation system and information
systems. Additional modules are considered for future development, for example
conversion of ZTD into IWV, animation plots or local monitoring.
176
J. Douša et al.
Fig. 3.101 Trop-Net modules – central (up) and disseminated (bottom)
3.7.1.1
Strategy for NRT Troposphere Monitoring
Several aspects are considered generally as important for developing the near realtime GNSS troposphere estimates: (a) high efficiency and low latency of GNSS
processing, (b) precise station coordinates fully consistent with troposphere estimates, and (c) robust system operated with a minimum manual interventions. For
this purpose, the Trop-NET system implemented three processing levels with intermediate solutions combined into a final solution, however, still efficient in near realtime fashion (Douša 2004a, b), Fig. 3.102:
1. Processing of small network clusters using a short-term data batch (yellow and
green).
2. Stacking of clusters in spatial domain into a single session network solution
(grey).
3. Stacking of network solutions in long-term solutions using previous solutions
(blue).
Originally, the Trop-NET system applied the 1-h session because of a limited
computer power in early 2000. In order to support global solution and for a reliable
integer ambiguity resolution at long baselines, the processing batch has been
extended to the 4-h session (Douša and Bennitt 2013). Hence, the level of processing
the original GNSS data remains redundant for 3 h as visible in Fig. 3.103. The
session network and sub-network solutions are temporarily saved in the form of
normal equations as intermediate products archived for next 30 days at least.
Various numbers of intermediate solutions are combined when estimating different parameters – tropospheric path delays, receiver coordinates or resolving integer
3 Advanced GNSS Processing Techniques (Working Group 1)
177
Fig. 3.102 Trop-NET processing in network clusters, spatial and temporal stacking solutions
Fig. 3.103 Trop-NET processing redundancy
phase ambiguities. Initially, all parameters are always estimated step-by-step using
the 4-h session when original GNSS observations are processed and precise satellite
positions introduced as known. First coordinates and tropospheric parameters are
then estimated from previous solutions of 1–2 days only before these are introduced
for resolving integer phase ambiguities. In later steps, the coordinates are estimated
from the time span of 28 days along with introducing integer phase ambiguities and
still estimating unresolved ambiguities as float values. Such coordinates are estimated within each individual hourly NRT solution and tied to the actual reference
datum. Such approach keeps the system free from any external process and provides
additional advantages: (a) all models being implicitly consistent for coordinate and
troposphere estimates, (b) coordinates are automatically updated within the system
178
J. Douša et al.
without need of an external information about their changes in time (implicitly
supports solutions in tectonically active areas such as Greece, Turkey and Iceland),
(c) any new station is configured once only with an implicit initialization of station
coordinates which is usually done from two past days. The tropospheric parameters
are finally estimated using the 12-h session when combining normal equations in
temporal domain. The ambiguities are estimated along with tropospheric parameters
in the final step because it guarantees more stable tropospheric product compared to
the ambiguities-fixed solution when using a short data session only.
Different parallelization strategies are used in various processing steps. The ‘predefinition’ regional clusters are used whenever necessarily applying a full correlation
model for the sub-network solution and such solutions are usually used also for
storing solution normal equations. On the other hand, adaptable strategy for parallel
processing uses optimal groups of sites or baselines, often suitable for autonomous
site or independent baseline processing. The examples are RINEX conversions,
pseudorange smoothing, receiver clock synchronization, ambiguity resolution etc.
The processing of regional clusters based on pre-defined configurations may still be
automatically adapted, e.g. merging too small clusters. The adaptation of clusters
uses several scenarios for generating optimal groups: (a) sorting station- or baselinespecific observation files, (b) using actually available stations, or (c) following
clusters from any previous step of the processing.
The processing system finally provides automatic warning and error messages
either via e-mail or via SMS indicating a temporary solution problem. A warning
message often informs about a temporary exclusion of station due to the incompatibility of the file header and the station metadata which always requires a station
manual reconfiguration. An error indicating a solution crash often represents a
temporary lack of data/products which is possibly within upcoming hours, or if
caused by a system-specific reason might need a manual intervention. The status of
processing solutions and monitoring indicators are archived along with the products.
3.7.2
Sofia University GNSS Analysis Centre (SUGAC): First
Processing Campaign
3.7.2.1
Motivation
T. Simeonov
Sofia University “St. Kliment Ohridski”, Sofia, Bulgaria
G. Guerova
Physics Faculty, Department of Meteorology and Geophysics, Sofia University
“St. Kliment Ohridski”, Sofia, Bulgaria
e-mail: [email protected]fia.bg
In Europe GNSS meteorology is a well-established field in both research and
operation, however, large regional differences were acknowledged in GNSS4SWEC
MoU, namely “while the production, exploitation and evaluation of operational
3 Advanced GNSS Processing Techniques (Working Group 1)
179
GNSS tropospheric products for NWP is well established in the Northern and
Western Europe, it is still an emerging R&D field in Eastern and South-Eastern
Europe”. In 2014, with the signature of national agreement between Sofia University
and Bulgarian BuliPOS GNSS network, the Sofia University GNSS Analysis Centre
(SUGAC, http://suada.phys.uni-sofia.bg/) was established to fill the gap of production of GNSS tropospheric products for Bulgaria and South-East Europe (Simeonov
et al. 2015). The GNSS4SWEC supported STSM for knowledge transfer and
processing of 1 year of tropospheric products from Bulgaria.
3.7.2.2
Main results
The first SUGAC processing campaign took place during the STSM of Tzvetan
Simeonov to University of Luxembourg (for details see Chap. 7). 7 stations from the
BULiPOS network (red pointers in Fig. 3.104) were processed using the NAvigation
Package for Earth Observation Satellites (NAPEOS) software version 3.3.1.
NAPEOS is developed and maintained by the European Space Operations Centre
of the European Space Agency (ESA). First SUGAC processing campaign was
performed using the Global Mapping Function and 10 elevation angle cutoff. The
RINEX files were processed using the PPP strategy and IGS satellite orbits and
clocks. ZTD was computed every 300 s (5 min) for one year – 2013. The ZTD data
was used to: (1) estimate IWV and (2) evaluate the numerical weather prediction
Fig. 3.104 Map of Bulgaria with marked (red dots) ground based stations of the BuliPOS GNSS
network used in first SUGAC processing campaign
180
J. Douša et al.
Fig. 3.105 Monthly mean IWV from GNSS (blue markers) and WRF (red markers) for: (a)
Burgas, (b) Shumen, (c) Stara Zagora, (d) Montana, (e) Varna and (f) Rozhen in 2013
(NWP) model for Bulgaria (Simeonov et al. 2016). In order to derive IWV with
sub-hourly temporal resolution the surface pressure and temperature from the
Weather Research and Forecast (WRF) NWP model were used. WRF model simulations were initialized at 00:00 UTC and computed on horizontal mesh of 9 km with
44 vertical levels over Bulgaria. Separately the IWV from WRF is computed by
integrating the vertical profile of the water vapour density. The comparisons of
monthly mean GNSS and WRF IWV at stations Burgas, Shumen, Stara Zagora,
Montana, Varna and Rozhen are presented in Fig. 3.105. At all stations, with
exception of Rozhen, the monthly mean IWV minimum is 10 [kg/m2] in December
2013 and the maximum is up to 25 [kg/m2] in June 2013. For station Burgas
(Fig. 3.105a) good agreement between the monthly mean IWV from GNSS and
WRF is seen with correlation coefficient between 0.96 and 0.84 and Root Mean
Square Error (RMSE) between 1.8 and 2.8 [kg/m2] (Fig. 3.106). The maximum and
minimum correlation is seen in winter and autumn, and spring and summer, respectively. Between stations Shumen (Fig. 3.105b) and Stara Zagora (Fig. 3.105c)
similarities in the IWV can be observed. The two stations also share low RMSE of
2.3 and 2.5 [kg/m2] respectively. For Shumen the lowest correlation is observed in
April and it remains low during the spring months. For Stara Zagora the correlation
coefficient stays low in summer with minimum from April till August. Montana
(Fig. 3.105d) is in Northwest Bulgaria where the influence of the Balkan mountains
is significant and the interaction with synoptic flows plays a major role for the IWV
distribution. The lowest GNSS and WRF IWV values are seen for December
12 [kg/m2] and the highest for June with 27 [kg/m2]. For Varna (Fig. 3.105e) of
interest is the difference between GNSS and WRF, which is seen in January–April
3 Advanced GNSS Processing Techniques (Working Group 1)
181
Fig. 3.106 IWV RMSE and correlation between model and GNSS datasets for the Bulipos
Network for 2013. Colors indicate: blue – first quarter of the year (JFM), green – second quarter
(AMJ), red – third quarter (JAS), orange – fourth quarter (OND)
and May–December period (marked with gray circle). From January to April the
IWV in the WRF is lower than the GNSS and from May to December it is the
opposite. Similar GNSS IWV jump between April and May is seen at Rozhen (gray
circle on Fig. 3.105f). The reason for IWV jump at station Varna is identified to be
change of antenna type in the RINEX file. However, for station Rozhen the reason
for the IWV change needs further investigation. A summary of the correlation and
RMSE for each season is presented in Fig. 3.106. For the cold part of the year
(January to March and October to December) the correlation between the model and
observation is highest and the RMSE is below 2.6 [kg/m2]. The warm part of the year
quarter 2 and 3 (April–September) is with RMSE over 2.6 [kg/m2] and correlation
below 0.9. This is expected and is related to the increased atmospheric dynamics and
summer time instability and convection, which are a well-known weakness of the
NWP models. A detailed investigation of WRF model performance in summer is
given in Chap. 4 of this report (Slavchev and Guerova).
3.7.2.3
Future Work
The first SUGAC processing campaign was a first step in building expertise in
Bulgaria with processing GNSS for remote-sensing the troposphere. The work will
continue by developing a pilot transnational severe weather service exploiting GNSS
182
J. Douša et al.
tropospheric products to enhance the safety, the quality of life and environmental
protection in the Balkan-Mediterranean region (Bulgaria, Cyprus and Greece).
3.7.3
TU Wien Near Real-Time GNSS Analysis Centre
in Austria (TUW AC): First Processing Results
G. Möller
Department of Geodesy and Geoinformation, TU Wien, Wien, Austria
e-mail: [email protected]
Since March 2017 TU Wien provides near real-time ZTDs for selected GNSS
reference sites in Austria and neighbouring countries, see Fig. 3.107.
The processing is based on dual-frequency GPS and GLONASS observations,
which are provided on a routine basis from the national reference network provider
EPOSA (www.eposa.at) in hourly batches. The processing is carried out at TU Wien
using the BSW52 double-difference processing strategy. The routines for processing
were established within the national research project GNSS-MET Austria in the
years 2009–2010 (see Karabatic et al. 2011) and were further refined during the
framework of the COST action. For reliable ambiguity resolution, the observations
available within the last 8 h are processed altogether. Therefore, it can be guaranteed
that at least 65% (long term average) of the ambiguities can be fixed to their integer
values. The accuracy of the tropospheric estimates is evaluated regularly at selected
IGS sites against the final IGS tropospheric estimates.
Figure 3.108 shows the results of the comparison, exemplary for station GRAZ
over the first 31 days in 2018. Therefore, from each near real-time solution only the
estimates of the last hour were considered. Except of a few outliers, the differences in
Fig. 3.107 GNSS station distribution. (Blue) GNSS stations of the Austrian reference network
EPOSA, (black) IGS and EUREF stations
3 Advanced GNSS Processing Techniques (Working Group 1)
183
Fig. 3.108 Comparison of TUW near real-time ZTDs with IGS final ZTDs at GNSS site Graz,
Austria. Analysed period: First 31 days in 2018
ZTD vary between +/ 1 cm with a mean bias of 1 mm and a standard deviation of
of about +/ 4 mm. A similar result is obtained for other IGS sites.
3.7.4
New Operational Solutions from ROB in Support
to Global NWP Models and Rapid-Update Numerical
Nowcasting
E. Pottiaux
Royal Observatory of Belgium, Brussels, Belgium
e-mail: [email protected]
Fostered by the successful developments done by ROB in the context of this COST
Action ES1206, the existing near real-time regional troposphere monitoring operated
continuously by ROB in the framework of E-GVAP was upgraded to the latest
BSW52, using now GPS + GLONASS observations, and processing them in a
double-difference batch approach with the latest modelling techniques. In that
184
J. Douša et al.
process, a special attention was given to develop a highly flexible and robust
processing chain using the BSW52 and the BPE (similar philosophy as the system
developed by GOP and described in Sect. 3.6.1), allowing thereby using the same
core processing system for various specific applications. This core systems also aims
to minimize the manual intervention.
As a consequence, we could extend our support to the meteorological community
by developing two new troposphere monitoring systems:
1. A 15-min updated regional troposphere monitoring to support nowcasting applications, and
2. A near real-time global troposphere monitoring to support global NWP models.
Similarly to the legacy ROB solution to E-GVAP, these two new contributions
uses the BSW52, GPS + GLONASS observations, the latest modelling techniques,
and are now fully operationally provided to all E-GVAP partners. These new
monitoring systems are shortly describes below.
3.7.4.1
New Sub-Hourly GPS + GLONASS Troposphere Monitoring
The sub-hourly operational solution operated by ROB includes about 235 GNSS
stations providing real-time observations throughout several NTRIP broadcaster
servers (Fig. 3.109). Its main objective is to enable rapid-update cycle NWP data
assimilation and non-numerical nowcasting applications in the BENELUX + UK
regions. This solution has an update cycle of 15 min, and is uploaded to E-GVAP
(named ROBQ) with a latency of max 10–15 min after the last observation included
in the processing. As expected due to the strong requirement of latency, the precision
of the solution is slightly less good than for a standard regional near real-time
solution, but it remains within the requirements imposed for such applications
(Offiler et al. 2010). Further developments and improvements are still expected to
improve this support.
Fig. 3.109 Left: Location of the GNSS stations currently included in ROB’s 15-min updated
operational contribution to E-GVAP. Right: ZTD time series (and its formal error below) for the
Belgian EPN station located in Denterghem from this solution (ROBQ, Status: 3 October 2017)
3 Advanced GNSS Processing Techniques (Working Group 1)
3.7.4.2
185
New Near Real-Time GPS + GLONASS Global Troposphere
Monitoring
The global operational solutions operated by ROB includes about 315 GNSS stations (named ROBG, Fig. 3.110). Its main objective is to support data assimilation in
global NWP models such as those from the U.K Met Office and Météo France. It
also potentially supports meteorological agencies outside Europe but collaborating
with E-GVAP (e.g. Environment Canada) to access our products. The precision of
the solution is similar to the one of the standard regional near real-time solution.
With ROB’s global solution, E-GVAP has now (solely) 3 global solutions (ROB,
GOP and U.K. Met Office). This allows redundancy and a combination process
(ASIC solution by ASI/e-Geos) at common GNSS sites, but ROBG also improves
the global coverage by processing sites that are not (yet) processed by the two other
ACs. Further works on this solution include performance tuning, and adding more
sites in specific area such as in Antarctica to further improve the spatial coverage for
global NWP models.
3.7.5
New Methods to User GNSS Vapor Estimates
for Meteorology (NUVEM)
R. Fernandes
University of Beira Interior, Covilhã, Portugal
e-mail: [email protected]
H. Valentim
University of Beira Interior, Covilhã, Portugal
e-mail: [email protected]
Fig. 3.110 Left: Location of the GNSS stations currently included in ROB’s hourly-updated global
operational contribution to E-GVAP (ROBG, Status: 3 October 2017). Right: ZTD time series (and
its formal error below) for the EPN station NYA1 located in Ny-Alesund, Norway
186
J. Douša et al.
P. Viterbo
Instituto Português do Mar e da Atmosfera, Lisbon, Portugal
e-mail: [email protected]
J. P. Martins
Instituto Português do Mar e da Atmosfera, Lisbon, Portugal
e-mail: [email protected]
A. Sá
Polytechnic Institute of Guarda, Guarda, Portugal
e-mail: [email protected]
J. Jones
Met Office, Exeter, UK
e-mail: jonathan.jones@metoffice.gov.uk
The goal was to include GNSS PWV estimates in weather forecast of Portugal,
especially in the decision process of warning dissemination of severe weather
situations. For that purpose, Instituto Português do Mar e da Atmosfera (IPMA)
and Space and Earth Analysis Laboratory (SEGAL) set up a scheme that provides
PWV based on GNSS estimates. Scripts were developed to automatically retrieve the
raw GNSS from ftp servers to the server at SEGAL each hour. NUVEM is using
146 stations from 6 GNSS networks over Portugal and Spain, see Fig. 3.111.
Fig. 3.111 Stations used to
estimate the PWV solutions
in the framework of
NUVEM project
3 Advanced GNSS Processing Techniques (Working Group 1)
3.7.5.1
187
GNSS Data Providers
Portugal: RENEP (1); SERVIR (2)
1. http://www.dgterritorio.pt/cartografia_e_geodesia/geodesia/redes_geodesicas/
renep/
2. http://www.igeoe.pt/servir/servir.asp
Spain: IGN (3); Castilla (4); Extremadura (5); Andalucia (6)
1.
2.
3.
4.
http://www.fomento.es
http://gnss.itacyl.es
http://194.224.247.162:8080/WebExtremadura/
http://www.juntadeandalucia.es/obraspublicasytransportes/
redandaluzadeposicionamiento/rap/
The product (GNSS-PWV maps) is available every 5 min with a 2 h delay and can be
checked through a dedicated website (http://nuvem.di.ubi.pt/) that was created for
the NUVEM project, where all relevant information, including operational results
were/are being published.
The ZTDs are estimated at SEGAL, which collects data provided by GNSS
networks and information about the GNSS satellite orbits provided by the NASA
Jet Propulsion Laboratory as inputs to the GIPSY-OASIS software that is responsible for calculating the ZTD. The conversion of ZTD to PWV is also performed at
SEGAL and requires additional information about the pressure and temperature in
the vicinity of the GNSS stations. In the developed scheme, these variables are
extracted and interpolated for each of the stations by IPMA from the forecasts
provided by the ECMWF and are sent to the SEGAL as soon as forecasts are
available, with up to about 12 h in advance of their use by the GIPSY. The PWV
estimates are then sent to the IPMA, where they are archived and made available to
the Operational Centre of Time Forecasting in the form of maps every 15 min,
although the available information allows maps every 5 min. In the moment, the
products currently available for free by JPL only allow 2 h delay estimates, limiting
their use in the context of nowcasting.
Although the project is over, SEGAL and IPMA still maintain the operation and
there are plans to improve the products (better outlier detection, quality flagging of
the retrievals, reducing delay, increasing availability, etc. as well as comparison to
other data sources).
3.7.6
Near Real-Time GNSS Processing at ASI/CGS, Italy
R. Pacione
e-GEOS/Centro di Geodesia Spaziale-Agenzia Spaziale Italiana, Matera, MT, Italy
e-mail: [email protected]
188
J. Douša et al.
ASI/CGS has been processing Near-Real Time data for E-GVAP since its beginning.
During the years of the COST Action ES1206, the existing Near Real-Time network
continuously analysed by ASI/CGS in the framework of E-GVAP was upgraded by
adding as many GNSS stations as possible in order to homogenize the coverage of
troposphere products over Italy.
GNSS data belonging to the following regional GNSS networks: Veneto, Liguria,
Piemonte, Friuli Venezia Giulia, Trentino, Umbria, Puglia, Calabria, Lazio,
Abruzzo, Campania and the NetGeo commercial GNSS network for the Sardinia
Island were added to the core network based on EPN and ASI stations. As of today,
about 250 stations in the Central Mediterranean part of Europe are recognized by the
NRT processing system.
In the E-GVAP framework, ASI/CGS is participating as Analysis Centre and acts
as Combination Centre delivering four tropospheric solutions:
1. Near Real Time ZTD (Operational, labelled ASI_): every hour, 150 ZTD estimates with a 1h450 latency for an European network of more than 250 sites
(Fig. 3.112, left);
2. Near Real Time Combined ZTD (Operational, labelled ASIC): every hour, the 150
ZTD estimates from the contributing E-GVAP Analysis Centres are combined
and made available to the project, using a combination scheme outlined in
Pacione et al. 2011. On hourly basis about 550 stations on a global scale are
combined (Fig. 3.112, right);
3. Near Real Time ZTD (Test, labelled ASIR): the aim of this solution is to evaluate
IGS RT products in hourly PPP for NWP application;
4. Sub-hourly ZTD (Test, labelled ASIS): the aim of this solution is to test RT
GNSS observation and products in sub-hourly PPP for now-casting application.
For ASI_, ASIR and ASIS solutions, GIPSY-OASIS II software (Webb and
Zumberge 1997) is used for data reduction. In particular, for ASI_ the standard
Fig. 3.112 Left: ASI E-GVAP operational network, Right: ASI E-GVAP Combined Network
3 Advanced GNSS Processing Techniques (Working Group 1)
189
technique of network adjustment is used fixing the IGS Ultra Rapid orbits. A 4-h
sliding window approach for data handling is applied with a sampling rate of 5 min
and an elevation cut-off angle for the data of 100. The ZWD is estimated every 5 min
with a stochastic model (random walk) and a constraint of 20 mm/h1/2. The station
coordinates are kept fixed to values provided by combining 1 month of daily postprocessed solutions and are updated every 30 days taking into account the tectonic
movements of the area as reported in Pacione and Vespe 2008.
3.7.7
New Analysis Centre (AUTh) and National Observatory
of Athens (NOA)
C. Pikridas
Aristotle University of Thessaloniki, Thessaloniki, Greece
e-mail: [email protected]
N. Zinas
Tekmon Geomatics, Ioánnina, Greece
e-mail: [email protected]
A. Ganas
National Observatory of Athens, Athens, Greece
e-mail: [email protected]
3.7.7.1
New Analysis Centre (AUTh)
In the frame of a Short Term Scientific Mission on October 2014, a new analysis
centre (AC) for near real-time GNSS tropospheric monitoring in Greece was
established at the Department of Surveying Engineering of the Aristotle University
of Thessaloniki (AUTh). Since then the AUTh Analysis Centre contributes to the
EGVAP hourly ZTDs from many permanent GNSS stations in Greece (Fig. 3.113)
using the Trop-NET Engine.
The AC provides a unique contribution of tropospheric products to the meteorological community for the E-GVAP project that cover the whole of Greece. During
the STSM the GOP’s (Geodetic Observatory Pecný) TropNET engine was installed.
AUTh AC operates the BSW52 and handles GNSS data from its own and collaborated networks. Additionally, 36 GNSS stations (Fig. 3.114) from IGS and EUREF
are included in the network processing scheme for datum definition and consistent
absolute tropospheric estimation.
The near real-time (NRT) processing engine includes the following three
modules:
1. flexible data, metadata, precise product and model downloading or mirroring,
190
Fig. 3.113 GNSS stations in Greece contributing to E-GVAP
Fig. 3.114 IGS and EUREF GNSS stations
J. Douša et al.
3 Advanced GNSS Processing Techniques (Working Group 1)
191
2. module for GNSS processing based on BSW52 and the BPE, and
3. tropospheric product filtering module for converting in the COST-716 format
version 2.2. Currently the tropospheric products are uploaded to the GOP data
centre, Met-Office and AUTh ftp archive.
In order to have a continuous quality monitoring of the estimated results (like
coordinates) and product evaluation the AUTh AC research team (GNSS_QC)
developed various scripts for automatic plots of ZTD and coordinates values for
each GNSS station (Fig. 3.115).
Finally, in collaboration with two other COST participating countries, Bulgaria
and Cyprus, the AUTh Research team received funding under the frame of the
European Territorial Cooperation Programme “Interreg V-B Balkan-Mediterranean
2014–2020” for the project BeRTISS (Balkan-Meditteranean Real Time Severe
weather Service).
The main objective of BeRTISS is to develop and establish a pilot transnational
severe weather service by exploiting Global Navigation Satellite Systems (GNSS)
tropospheric products to enhance the quality and safety of life in the BalkanMediterranean region. This monitoring service will provide continuous and
uninterruptible information for nowcasting, forecasting and early warning for
Fig. 3.115 Coordinate residuals for GNSS reference station AUT1
192
J. Douša et al.
PWV using the GNSS derived tropospheric products and WRF (Weather Research
and Forecasting) model that will be tangible and visible to the public through a
dedicated web-platform. In detail, the aims of the project are: (1) Integration of
networks of GNSS stations located in the three countries in a unified system,
(2) Collection, processing and analysis of GNSS observations and tropospheric
products, (3) Calculation of the meteorological parameter IWV/PWV for more
accurate short-term prediction of severe weather events and (4) Creation of a
dedicated website to provide in near real-time the National Meteorological Services
and the public with PWV data and warnings of severe weather events.
BeRTISS comprises the continuation of the EU-COST Action “GNSS4SWEC”
in particular with respect to the expansion of GNSS tropospheric products in one of
the Europe’s most remote region and vulnerable to climate change.
3.7.7.2
National Observatory of Athens (NOA)
National Observatory of Athens (NOA) operates NOANET the nationwide geodetic
network of 22 CORS stations (www.gein.noa.gr/gps.html). NOANET daily 30-s
data are distributed via the GSAC web service http://194.177.194.238:8080/
noanetgsac/ as part of the ongoing project EPOS-IP https://epos-ip.org/. In total
the NOA GSAC distributes data from 62 CORS stations in SE Europe. In addition,
NOA (1) conducted several GPS field campaigns (re-measuring the position of
benchmarks) of the CRL/Lefkada/Messinia network in scheduled missions (2) continued installation and maintenance of CORS stations, in the CRL area, in Rhodes
(station KATC owned by UNAVCO) and in Messinia (new station ANIK) and
(3) conducted installation and maintenance of six (6) continuous GNSS stations in
the Ionian Sea area after the 17 November 2015 Earthquake in Lefkada (DRAN,
EXAN, ASSO, FISK, KIPO, VLSM).
NOA continues to apply space geodesy techniques (SAR interferometry and
GNSS) as an important tool for mapping regional surface deformations due to
tectonic motions and large earthquakes (e.g. Ganas et al. 2015; 2016; 2017; ongoing
research by Athanassios Ganas, Panagiotis Elias, Panagiotis Argyrakis and
Alexandra Moshou). In addition, NOA is engaged in research activities within the
CRL project (http://crlab.eu/) with emphasis on the effect of the troposphere (ongoing research by Nikos Roukounakis and Panagiotis Elias). The troposphere introduces a path delay in the radio signal, which, in the case of GPS, can be partially
removed with the use of specialized mapping functions. Moreover, tropospheric
stratification and short wavelength spatial turbulences produce an additive noise to
the ground deformation calculated by the (multitemporal) INSAR methodology. The
objective is to further correct the vertical component in GPS measurements with the
use of a high resolution meteorological model (WRF), producing a 3D tomography
of the troposphere. Thus, the knowledge of the tropospheric parameters along the
propagation medium can be used to estimate and minimize the effect of this noise, so
that the remaining signal represents the deformation mostly due to tectonic or other
geophysical processes (Roukounakis et al. 2015). The data-contributing stations
3 Advanced GNSS Processing Techniques (Working Group 1)
193
belong to the Corinth Rift Laboratory and NOANET networks, which monitor the
seismicity of the region on a permanent basis. Results are compared with tropospheric delays derived from WRF re-analysis.
3.7.8
New Analysis Centre in Iceland, Icelandic
Meteorological Office (IMO)
S. Thorsteinsson
The Icelandic Meteorological Institute, Reykjavík, Iceland
e-mail: [email protected]
B. G. Ófeigsson
The Icelandic Meteorological Institute, Reykjavík, Iceland
e-mail: [email protected]
Icelandic Meteorological Office (IMO) operates most monitoring networks of natural hazards in Iceland, see Fig. 3.116. Its operations range from Meteorological
monitoring, Hydrological monitoring to volcano and seismic hazard monitoring. As
Fig. 3.116 Continuous GNSS network in Iceland (ISGPS). The diamonds mark the location of a
GNSS station
194
J. Douša et al.
a part of this monitoring effort IMO operates a continuous GNSS network mostly
used for volcano monitoring.
With the support of Geodetic Obseratory Pecný (GOP), IMO established an
analysis centre for near real-time regional troposphere monitoring, using GOP’s
Trop-NET system in March 2016. Since then the majority of Icelandic continuous
GNSS stations already operated by IMO are routinely processed. The analyses is
now a part of IMO’s continuous operational systems where it has become one of the
analysis centres providing GNSS ZTD data to E-GVAP.
IMO has started and plans to continue to do assimilation impact studies with the
GNSS ZTD data as well as from 4 GNSS sites in Greenland gotten from E-GVAP in
HARMONIE on the 2.5 km IGB grid domain. New decision regarding the common
operational system between IMO and the Danish Meteorological Institute (DMI) is
to extend the domain to cover the whole Greenland and Iceland and its surrounding
islands, termed IGB domain. The HARMONIE tools and the Icelandic processing
GNSS ZTD centre that we have developed in COST ES1206 to monitor convective
clouds and severe weather conditions will become useful for IMO.
3.7.9
New Analysis Centre in Hungary (BUTE)
Sz. Rozsa
Budapest University of Technology and Economics, Budapest, Hungary
e-mail: [email protected]
The new Analysis Centre (SGO1) was set up in 2014. It processes the Hungarian
active GNSS network (37 stations with the mean distance of 60 kms) including some
permanent stations from the neighbouring countries (19) as well as some EUREF
station as fiducial stations. The near realtime processing is done using BSW52 and
the estimated ZTD values are automatically transmitted to the E-GVAP programme.
References
Alber, C., Ware, R. H., Rocken, C., & Solheim, F. S. (1997). GPS surveying with 1 mm precision
using corrections for atmospheric slant path delays. Geophysical Research Letters, 24,
1859–1862.
Alber, C., Ware, R., Rocken, C., & Braun, J. (2000). Obtaining single path phase delays from GPS
double differences. Geophysical Research Letters, 27, 2661–2664. https://doi.org/10.1029/
2000gl011525.
Altiner, Y., Söhne, W., & Weber, G. (2009). Echtzeitnahe Schätzung troposphärischer LaufzeitVerzögerung und Koordinatenüberwachung, AVN 2/2009, pp. 42–47.
Altiner, Y., Mervart, L., Söhne, W., & Weber, G. (2010). Real-time PPP results from global orbit
and clock corrections. https://igsac-cnes.cls.fr/documents/egu10/EGU2010-11969_Altiner.pdf
Altiner, Y., Söhne, W., & Weber, G. (2011). Von NRT zu RT. http://www.uni-stuttgart.de/gi/
research/Geodaetische_Woche/2011/Session4/S4-18-Altiner.pdf
3 Advanced GNSS Processing Techniques (Working Group 1)
195
Askne, J., & Nordius, H. (1987). Estimation of tropospheric delay for microwaves from surface
weather data. Radio Science, 22(3), 379–386.
Bar-Sever, Y. E., & Kroger, P. M. (1998). Estimating horizontal gradients of tropospheric path
delay with a single GPS receiver. Journal of Geophysical Research, 103, 5019–5035.
Bennitt, G., & Jupp, A. (2012). Operational assimilation of GPS zenith total delay observations into
the UK met Office numerical weather prediction models. Monthly Weather Review, 140(8),
2706–2719. https://doi.org/10.1175/MWR-D-11-00156.1.
Blewitt, G. (1989). Carrier phase ambiguity resolution for the global positioning system applied to
geodetic baselines up to 2000 km. Journal of Geophysical Research, 94, 10,187–10,203. https://
doi.org/10.1029/JB094iB08p10187.
Boehm, J., Niell, A., Tregoning, P., & Schuh, H. (2006a). Global mapping function (GMF): A new
empirical mapping function based on numerical weather model data. Geophysical Research
Letters, 33, L07304. https://doi.org/10.1029/2005GL025546.
Boehm, J., Werl, B., & Schuh, H. (2006b). Troposphere mapping functions for GPS and very long
baseline interferometry from European Centre for Medium-Range Weather Forecasts operational analysis data. Journal of Geophysical Research, 111, B02406. https://doi.org/10.1029/
2005JB003629.
Böhm, J., Möller, G., Schindelegger, M., Pain, G., & Weber, R. (2015). Development of an
improved empirical model for slant delays in the troposphere (GPT2w). GPS Solutions,
Springer, 19(3), 433–441.
Brenot, H., & Warnant, R. (2008). Characterization of the tropospheric small-scale activity,
Technical Report ESA, WP250, GALOCAD project. https://orbi.uliege.be/jspui/bitstream/
2268/84849/1/GALOCAD-WP250-Report-OUT250-1_v2.pdf
Brenot, H., Ducrocq, V., Walpersdorf, A., Champollion, C., & Caumont, O. (2006). GPS zenith
delay sensitivity evaluated from high-resolution numerical weather prediction simulations of the
8–9 September 2002 flash flood over southeastern France. Journal of Geophysical Research,
111, D15105. https://doi.org/10.1029/2004JD005726.
Brenot, H., Neméghaire, J., Delobbe, L., Clerbaux, N., De Meutter, P., Deckmyn, A., Delcloo, A.,
Frappez, L., & Van Roozendael, M. (2013). Preliminary signs of the initiation of deep
convection by GNSS. Atmospheric Chemistry and Physics, 13, 5425–5449. https://doi.org/10.
5194/acp-13-5425-2013 (licensed under CC BY 3.0, https://creativecommons.org/licenses/by/
3.0/).
Brenot, H., Wautelet, G., Warnant, R., Nemeghaire, J., & Van Roozendael, M. (2014a) GNSS
meteorology and impact on NRT position. Proceedings of the European Navigation Conference
(ENC), Rotterdam, The Netherlands.
Brenot, H., Walpersdorf, A., Reverdy, M., van Baelen, J., Ducrocq, V., Champollion, C., Masson,
F., Doerflinger, E., Collard, P., & Giroux, P. (2014b). A GPS network for tropospheric
tomography in the framework of the Mediterranean hydrometeorological observatory
Cévennes-Vivarais (southeastern France). Atmospheric Measurement Techniques, 7, 553–578.
https://doi.org/10.5194/amt-7-553-2014.
Caissy, M., Agrotis, L., Weber, G., Hernandez-Pajares, M., & Hugentobler, U. (2012).
INNOVATION-coming soon-the international GNSS real-time service. GPS World, 23, 52–58.
Champollion, C., Masson, F., Van Baelen, J., Walpersdorf, A., Chéry, J., & Doerflinger, E. (2004).
GPS monitoring of the tropospheric water vapour distribution and variation during the
September 9, 2002, torrential precipitation episode in the Cévennes (southern France). Journal
of Geophysical Research, 109, D24.
Chen, G., & Herring, T. A. (1997). Effects of Atmospheric Azimuthal Asymetry on the Analysis of
Space Geodetic Data. Geophysical Research Letters, 102(20), 489–420. 502.
Dach, R., & Jean, Y. (2015). International GNSS service, technical report 2014. Pasadena: IGS
Central Bureau.
Dach, R., Schaer, S., Lutz, S., Baumann, C., Bock, H., Orliac, E., Prange, L., Thaller, D., Mervart,
L., Jäggi, A., Beutler, G., Brockmann, E., Ineichen, D., Wiget, A., Weber, G., Habrich, H.,
Söhne, W., & Ihde, J. (2014). Steigenberger, P., and Hugentobler, U: CODE IGS Analysis
196
J. Douša et al.
Center Technical Report 2013, Dach, R., & Jean, Y. (Eds.), IGS 2013 Technical Report,
pp. 21–34.
Dach, R., Lutz, S., Walser, P., & Fridez, P. (2015). Bernese GNSS Software Version 5.2, University
of Bern, Bern Open Publishing. https://doi.org/10.7892/boris.72297.
Dach, R., Stefan, S., Arnold, D., Orliac, E., Prange, L., Suśnik, A., Villiger, A., & Jäggi, A. (2016)
CODE final product series for the IGS, Published by Astronomical Institute, University of Bern.
https://doi.org/10.7892/boris.75876.
Davis, J. L., Herring, T. A., Shapiro, I. I., Rogers, A. E. E., & Elgered, G. (1985). Geodesy by
interferometry: Effects of atmospheric modeling errors on estimates of baseline length. Radio
Science, 20, 1593–1607.
Davis, J. L., Elgered, G., Niell, A. E., & Kuehn, C. E. (1993). Groundbased measurements of
gradients in the “wet” radio refractivity of air. Radio Science, 28, 1003–1018.
De Haan, S. (2006). National/regional operational procedures of GPS water vapour networks and
agreed international procedures, Rep WMO/TD-No, 1340:20, KNMI, Netherlands.
Dee, D. P., et al. (2011). The ERA-Interim reanalysis: Configuration and performance of the data
assimilation system. Quarterly Journal of the Royal Meteorological Society, 137(656),
553–597.
Delobbe, L., & Holleman, I. (2006). Uncertainties in radar echo top heights used for hail detection.
Meteorological Applications, 13, 361–374.
Delrieu, G., Nicol, J., Yates, E., Kirstetter, P.-E., Creutin, J.-D., Anquetin, S., Obled, C., Saulnier,
G.-M., Ducrocq, V., Gaume, E., Payrastre, O., Andrieu, H., Ayral, P.-A., Bouvier, C., Neppel,
L., Livet, M., Lang, M., Parent du-Châtelet, J., Walpersdorf, A., & Wobrock, W. (2005). The
Catastrophic Flash-Flood Event of 8–9 September 2002 in the Gard Region, France: a First Case
Study for the Cévennes-Vivarais Mediterranean Hydrometeorological Observatory, Journal of
Hydrometeorology, 6, 34–51. https://doi.org/10.1175/JHM400.1.
Deng, Z., Schöne, T., & Gendt, G. (2014). Status of the TIGA Tide Gauge Data Reprocessing at
GFZ. International Association of Geodesy Symposia, October 2014.
Deng, Z., Uhlemann, M., Fritsche, M., Dick, G., & Wickert, J. (2015). Troposphere parameters
derived from multi-GNSS data processing at GFZ. Wien, EGU 2015.
Deng, Z., Nischan, T., & Bradke, M. (2017). Multi-GNSS Rapid Orbit-, Clock- & EOP-Product
Series. GFZ Data Services. https://doi.org/10.5880/GFZ.1.1.2017.002.
Dick, G., Gendt, G., & Reigber, C. (2001). First experience with near real-time water vapor
estimation in a German GPS network. Journal of Atmospheric and Solar – Terrestrial Physics,
63, 1295–1304. https://doi.org/10.1016/S1364-6826(00)00248-0.
Ding, W., Teferle, F. N., Kazmierski, K., Laurichesse, D., & Yuan, Y. (2017). An evaluation of
real-time troposphere estimation based on GNSS Precise Point Positioning. Journal of Geophysical Research Atmosphere, 122(5), 2779–2790. https://doi.org/10.1002/2016JD025727.
Dong, D.-N., & Bock, Y. (1989). Global positioning system network analysis with phase ambiguity
resolution applied to crustal deformation studies in California. Journal of Geophysical
Research, 94, 3949–3966. https://doi.org/10.1029/JB094iB04p03949.
Douša, J. (2001a). Towards an operational near-real time precipitable water vapor estimation.
Physics andChemistry of the Earth, Part. A, 26, 189–194. https://doi.org/10.1016/S1464-1895
(01)00045-X.
Douša, J. (2001b). The impact of ultra-rapid orbits on precipitable water vapor estimation using
ground GPS network. Physics and Chemistry of the Earth, 26(6–8), 393–398.
Douša, J. (2004a). Evaluation of tropospheric parameters estimated in various routine analysis.
Physics and Chemistry of the Earth, 29(2–3), 167–175.
Douša, J. (2004b). Precise orbits for ground-based GPS meteorology: Processing strategy and
quality assessment of the orbits determined at geodetic observatory pecný. Journal of the
Meteorological Society of Japan, 82, 371–380.
Douša, J. (2010). Precise near real-time GNSS analyses at geodetic observatory pecný – Precise
orbit determination and water vapour monitoring. Acta Geodynamics et Geomaterialia, 7(157),
1–11.
3 Advanced GNSS Processing Techniques (Working Group 1)
197
Douša, J. (2012). Developments of the GLONASS ultra-rapid orbit determination at geodetic
observatory pecný. In S. Kenyon, M. C. Pacino, & U. Marti (Eds.), Geodesy of planet earth
(IAG Symposia Series) (Vol. 136, pp. 1029–1036). Dordrecht: Springer.
Douša, J., & Bennitt, G. V. (2013). Estimation and evaluation of hourly updated global GPS zenith
Total delays over ten months. GPS Solutions, 17, 453–464. https://doi.org/10.1007/s10291-0120291-7.
Douša, J., & Eliaš, M. (2014). An improved model for calculating tropospheric wet delay.
Geophysical Research Letters, 41, 4389–4397.
Douša, J., & Souček, P. (2005). The results of near real-time COST-716 GPS campaign from
geodetic observatory Pecný, In J. Sledzinski (Eds.), Proceedings of the EGU G9 symposium
(pp. 139–150). Reports on Geodesy, Warsaw University of Technology, Institute of Geodesy
and Geodetic Astronomy, Warsaw, Poland.
Douša, J., & Václavovic, P. (2014). Real-time zenith tropospheric delays in support of numerical
weather prediction applications. Advances in Space Research, 53(9), 1347–1358. https://doi.
org/10.1016/j.asr.2014.02.021.
Douša, J., & Václavovic, P. (2016). Evaluation of ground-based GNSS tropospheric products at
Geodetic Observatory Pecný. Proceedings of the IAG Symposia Series, Springer, Rizos Ch. &
Willis P. (Eds.). 143:759–766. 10.1007/1345_2015_157.
Douša, J., Elias, M., Veerman H., van Leeuwen, S. S., Zelle, H., de Haan, S., Martellucci, A., &
Perez, R. O. (2015a). High accuracy tropospheric delay determination based on improved
modelling and high resolution numerical weather model. Proceedings of the ION GNSS 2015,
Institute of Navigation, Tampa, Florida, USA, September 14–18, pp. 3734–3744.
Douša, J., Václavovic, P., Krč, P., Eliaš, M., Eben, E., & Resler, J. (2015b). NWM forecast
monitoring with near real-time GNSS products, In Proceedings of the 5th scientific Galileo
Colloquium, Braunschweig, Germany, October 27–29. http://old.esaconferencebureau.com/
docs/default-source/15a08_session1/044-dousa.pdf?sfvrsn¼2
Douša, J., Dick, G., Kačmařík, M., Brožková, R., Zus, F., Brenot, H., Stoycheva, A., Möller, G., &
Kaplon, J. (2016). Benchmark campaign and case study episode in Central Europe for development and assessment of advanced GNSS tropospheric models and products. Atmospheric
Measurement Techniques, 9, 2989–3008. https://doi.org/10.5194/amt-9-2989-2016. (licensed
under CC BY 3.0, https://creativecommons.org/licenses/by/3.0/).
Douša, J., Václavovic, P., & Eliaš, M. (2017). Tropospheric products of the second European GNSS
reprocessing (1996–2014). Atmospheric Measurement Techniques, 10, 1–19. https://doi.org/10.
5194/amt-10-1-2017. (licensed under CC BY 3.0, https://creativecommons.org/licenses/by/3.0/).
Douša, J., Václavovic, P., Zhao, L., & Kačmařík, M. (2018a). New adaptable all-in-one strategy for
estimating advanced tropospheric parameters and using real-time orbits and clocks. Remote
Sensing, 10, 232. https://doi.org/10.3390/rs10020232. (licensed under CC BY 4.0, https://
creativecommons.org/licenses/by/4.0/).
Douša, J., Eliaš, M., Václavovic, P., Eben, K., & Krc, P. (2018b). A two-stage tropospheric
correction combining data from GNSS and numerical weather model. GPS Solutions, 22, 77.
https://doi.org/10.1007/s10291-018-0742-x.
Dow, J. M., Neilan, R. E., & Rizos, C. (2009). The international GNSS service in a changing
landscape of global navigation satellite systems. Journal of Geodesy, 83, 191–198. https://doi.
org/10.1007/s00190-008-0300-3.
Elgered, G., Davis, J. L., Herring, T. A., & Shapiro, I. I. (1991). Geodesy by radio interferometry:
Water vapour radiometry for estimation of the wet delay. Journal of Geophysical Research, 96,
6541–6555.
Elósegui, P., Davis, J. L., Jaldehag, R. T. K., Niell, A. E., & Shapiro, I. I. (1995). Geodesy using the
global positioning system: The effects of signal scattering on estimates of site position. Journal
of Geophysical Research, 100, 99219934.
Elósegui, P., Davis, J. L., Gradinarsky, L. P., Elgered, G., Johansson, J. M., Tahmoush, D. A., &
Ruis, A. (1999). Sensing atmospheric structure using small-scale space geodetic networks.
Geophysical Research Letters, 26, 2445–2448.
198
J. Douša et al.
Flores, A., Ruffini, G., & Rius, A. (2000). 4D tropospheric tomography using GPS slant wet delays.
Annales de Geophysique, 18, 223–234. https://doi.org/10.1007/s00585-000-0223-7.
Ge, M., Gendt, G., Dick, G., Zhang, F. P., & Rothacher, M. (2006). A new data processing strategy
for huge GNSS global networks. Journal of Geodesy, 80(4), 199–203.
Gendt, G., Dick, G., Reigber, C., Tomassini, M., Liu, Y., & Ramatschi, M. (2004). Near real time
GPS water vapor monitoring for numerical weather prediction in Germany. Journal of the
Meteorological Society of Japan, 82, 361–370.
Gradinarsky, L. P. (2002). Sensing atmospheric water vapor using radio waves, Ph.D. thesis,
School of Electrical Engineering, Chalmers University of Technology, Göteborg, Sweden.
Guerova, G., Jones, J., Douša, J., Dick, G., de Haan, S., Pottiaux, E., Bock, O., Pacione, R., Elgered,
G., Vedel, H., & Bender, M. (2016a). Review of the state of the art and future prospects of the
ground-based GNSS meteorology in Europe. Atmospheric Measurement Techniques, 9,
5385–5406. https://doi.org/10.5194/amt-9-5385-2016.
Guerova, G., Simeonov, Tzv.,& Vassileva, K. (2016b). Comparison of GNSS tropospheric products obtained by two processing strategies. In Proceedings of international SYMPOSIUM on
modern technologies, education and professional practice in geodesy and related fields. Sofia,
Bulgaria, 3–4/11/2016, 8 p. ISSN 2367-6051. Available at: http://suada.phys.uni-sofia.bg/
wordpress/wp-content/uploads/2017/07/Guerova_etal_Sofia-Symposium_2016.pdf
Hadaś, T., & Bosy, J. (2015). IGS RTS precise orbits and clocks verification and quality degradation over time. GPS Solutions, 19, 93–105. https://doi.org/10.1007/s10291-014-0369-5.
Hadaś, T., Teferle, F. N., Kaźmierski, K., Hordyniec, P., & Bosy, J. (2017). Optimum stochastic
modeling for GNSS tropospheric delay estimation in real-time. GPS Solutions, 21(3),
1069–1081. Berlin – Heidelberg.
Herring, T. A. (1992). Modeling atmospheric delays in te analysis of space geodetic data. In
Proceedings of the symposium on refraction of transatmospheric signals in geodesy. Netherlands Geodetic Commission, Publications on Geodesy, Delft, the Netherlands.
Herring, T. A., King, R. W., & McClusky, S. C. (2010). Documentation for the GAMIT GPS
analysis software, version 10.4, Technical report. Cambridge: Massachusetts Institute of
Technology.
Hofmeister, A., & Böhm, J. (2017). Application of ray-traced tropospheric slant delays to geodetic
VLBI analysis. The Journal of Geodesy, 91(8), 945–964.
Huisman L., van der Marel H., & Teunissen P. (2009). CORS local-site finger-printing using
undifferenced least squares GNSS phase residuals. In International global navigation satellite
systems society IGNSS symposium 2009, Holiday Inn, Surfers Paradise, Qld, Australia,
December 1–3
Hunegnaw, A., Teferle, F. N., Bingley, R., & Hansen, D. (2015). Status of TIGA activities at the
British Isles continuous GNSS facility and the University of Luxembourg, C. Rizos & P. Willis
(Eds.), International Association of Geodesy Symposia, 143:617–623. https://doi.org/10.1007/
1345-2015-77. Cham: Springer.
Hunegnaw, A., Teferle, F. N., Abraha, K. E., Santamaría-Gómez, A., Gravelle, M., Wöppelman,
G., Schöne, T., Deng, Z., Bingley, R., Hansen, D., Sanchez, L., & Moore, M. (2017). On the
Scientific Applications of IGS Products: An Assessment of the Reprocessed TIGA Solutions and
Combined Products. IGS Workshop 2017, Paris.
Iwabuchi, T., Miyazaki, S., Heki, K., Naito, I., & Hatanaka, Y. (2003). An impact of estimating
tropospheric delay gradients on tropospheric delay estimations in the summer using the Japanese nationwide GPS array. Journal of Geophysical Research, 108, 4315. https://doi.org/10.
1029/2002JD002214.
Kačmařík, M., Douša, J., Dick, G., Zus, F., Brenot, H., Möller, G., Pottiaux, E., Kapłon, J.,
Hordyniec, P., Václavovic, P., & Morel, L. (2017). Inter-technique validation of tropospheric
slant total delays. Atmospheric Measurement Techniques, 10, 2183–2208. https://doi.org/10.
5194/amt-10-2183-2017. (licensed under CC BY 3.0, https://creativecommons.org/licenses/by/
3.0/).
3 Advanced GNSS Processing Techniques (Working Group 1)
199
Kapłon, J., Hordyniec, P., & Rohm, W. (2017). Analysis of systematic effects in slant total delay
estimation with PPP. Presentation G06-1-06 at IAG-IASPEI 2017 Joint Scientific Assembly of
the International Association of Geodesy (IAG) and International Association of Seismology
and Physics of the Earth’s Interior (IASPEI), Kobe, Japan, August 1.
Karabatic, A., Weber, R., & Haiden, T. (2011). Near real-time estimation of tropospheric water
vapour content from ground based GNSS data and its potential contribution to weather
now-casting in Austria. Advances in Space Research, 47(10), 1691–1703. https://doi.org/10.
1016/j.asr.2010.10.028.
King, R. W., Masters, E. G., Rizos, C., Stolz, A., & Collins, J. (1985). Surveying with GPS,
Monograph 9, School of Surveying, University of New South Wales, Kensington, Australia.
Klos, A., Hunegnaw, A., Teferle, F. N., Abraha, K. E., Ahmed, F., & Bogusz, J. (2018). Statistical
significance of trends in Zenith Wet Delay from re-processed GPS solutions. GPS Solutions, 22,
51. https://doi.org/10.1007/s10291-018-0717-y.
Landskron, D., & Böhm, J. (2017). VMF3/GPT3: Refined discrete and empirical troposphere
mapping functions. The Journal of Geodesy. https://doi.org/10.1007/s00190-017-1066-2.
(licensed under CC BY 4.0, https://creativecommons.org/licenses/by/4.0/).
Laurichesse, D. (2011). The CNES real-time PPP with undifferenced integer ambiguity resolution
demonstrator. In Proceedings of the ION GNSS 2011, Portland, Oregon.
Laurichesse, D., Cerri, L., Berthias, J. P., & Mercier, F. (2013). Real time precise GPS constellation
and clocks estimation by means of a Kalman filter. In Proceedings ION GNSS 2013
(pp. 1155–1163), Institute of Navigation, Nashville, Tennessee, USA, September 16–20.
Leick, A. (1989). GPS satellite surveying. New York: Wiley-Interscience.
Leick, A. (2004). GPS satellite surveying (3rd ed., 435p). New York: Wiley.
MacMillan, D. S. (1995). Atmospheric gradients from very long baseline interferometry observations. Geophysical Research Letters, 22(9), 97–102. https://doi.org/10.1029/95GL00887.
Marini, J. W. (1972). Correction of satellite tracking data for an arbitrary tropospheric profile. Radio
Science, 7(2), 223–231.
Möller, G. (2017) Reconstruction of 3D wet refractivity fields in the lower atmosphere along
bended GNSS signal paths, Dissertation, TU Wien, Department of Geodesy and
Geoinformation. http://repositum.tuwien.ac.at/obvutwhs/content/titleinfo/2268559
Möller, G., Weber, R., & Böhm, J. (2014). Improved troposphere blind models based on numerical
weather data. NAVIGATION: Journal of the Institute of Navigation, 61(3), 203–211.
Montenbruck, O., Steigenberger, P., Prange, L., Deng, Z., Zhao, Q., Perosanz, F., Romero, I., Noll,
C., Stürze, A., Weber, G., Schmid, R., MacLeod, K., & Schaer, S. (2017). The Multi-GNSS
Experiment (MGEX) of the International GNSS Service (IGS) – Achievements, prospects and
challenges. Advances in Space Research, 59, 1671–1697. https://doi.org/10.1016/j.asr.2017.01.
011.
Morel, L., Pottiaux, E., Durand, F., Fund, F., Boniface, K., de Oliveira, P. S., & Van Baelen,
J. (2014). Validity and behaviour of tropospheric gradients estimated by GPS in Corsica.
Advances in Space Research. https://doi.org/10.1016/j.asr.2014.10.004.
Niell, A. E. (1996). Global mapping functions for the atmosphere delay at radio wavelengths.
Journal of Geophysical Research, 101, 3227–3246. https://doi.org/10.1029/95JB03048.
Ning, T., Wang, J., Elgered, G., Dick, G., Wickert, J., Bradke, M., & Sommer, M. (2016). The
uncertainty of the atmospheric integrated water vapour estimated from GNSS observations.
Atmospheric Measurement Techniques, 9, 79–92. https://doi.org/10.5194/amt-9-79-2016.
Offiler, D. et al. (2010). EIG EUMETNET GNSS Water Vapour Programme (E-GVAP-II), Product
Requirements Document, Version 1.0 – 21 December 2010.
Pacione, R., & Vespe, F. (2008). Comparative studies for the assessment of the quality of near-real
time GPS-derived atmospheric parameters. Journal of Atmospheric and Oceanic Technology,
25, 701–714.
Pacione, R., Pace, B., de Haan, S., Vedel, H., Lanotte, R., & Vespe, F. (2011). Combination
methods of tropospheric time series. Advances in Space Research, 47, 323–335. https://doi.org/
10.1016/j.asr.2010.07.021.
200
J. Douša et al.
Pacione, R., Araszkiewicz, A., Brockmann, E., & Douša, J. (2017). EPN-Repro2: A reference
GNSS tropospheric data set over Europe. Atmospheric Measurement Techniques, 10,
1689–1705. https://doi.org/10.5194/amt-10-1689-2017.
Petit, G., Luzum, B. (red.), IERS Conventions. (2010). IERS Technical Note No. 36, Frankfurt am
Main: Verlag des Bundesamts für Kartographie und Geodäsie, ISBN 3-89888-989-6
Pottiaux, E., Brockmann, E., Söhne, W., & Bruyninx, C. (2009). The EUREF – EUMETNET
collaboration: First experience and potential benefits. Bulletin of Geodesy and Geomatics, 3,
269–288.
Prange, L., Orliac, E., Dach, R., Arnold, D., Beutler, G., Schaer, S., & Jäggi, A. (2017). CODE’s
five-system orbit and clock solution—The challenges of multi-GNSS data analysis. Journal of
Geodesy, 91, 345–360. https://doi.org/10.1007/s00190-016-0968-8.
Rebischung, P., Griffiths, J., Ray, J., Schmid, R., Collilieux, X., & Garayt, B. (2012). IGS08: The
IGS realization of ITRF2008. GPS Solutions, 16, 483494. https://doi.org/10.1007/s10291-0110248-2.
Rebischung, P., Altamimi, Z., Ray, J., & Garayt, B. (2016). The IGS contribution to ITRF2014.
Journal of Geodesy, 90, 611–630. https://doi.org/10.1007/s00190-016-0897-6.
Roukounakis, N., et al. (2015). Improvement of the vertical component of GPS and INSAR
measurements in the western Corinth Gulf (Greece), by the use of high-resolution meteorological modeling of the lower troposphere: The PaTrop Experiment, 12th European Conference on
Applications of Meteorology (ECAM), EMS2015-544;11 September 2015, Sofia, Bulgaria.
Ruffini, G., Kruse, L. P., Rius, A., Bürki, B., & Cucurull, L. (1999). Estimation of tropospheric
zenith delay and gradients over the Madrid area using GPS and WVR data. Geophysical
Research Letters, 26(4), 447–450. https://doi.org/10.1029/1998GL900238.
Saastamoinen, J. (1972). Atmospheric correction for the troposphere and stratosphere in radio
ranging of satellites. The Geophysical Monograph Series, 15, 247–251.
Schmid, R., Dach, R., Collilieux, X., Jäggi, A., Schmitz, M., & Dilssner, F. (2016). Absolute IGS
antenna phase center model igs08.atx: status and potential improvements. Journal of Geodesy,
90(4), 343–364.
Seeber, G. (2003). Satellite geodesy (2nd ed., 589p). New York: de Gruyter.
Shoji, Y., Nakamura, H., Iwabuchi, T., Aonashi, K., Seko, H., Mishima, K., Itagaki, A., Ichikawa,
R., & Ohtani, R. (2004). Tsukuba GPS dense net campaign observation: Improvement in GPS
analysis of slant path delay by stacking one-way post phase residuals. Journal of the Meteorological Society of Japan, 82(1B), 301–314. https://doi.org/10.2151/jmsj.2004.301.
Simeonov Tzv., Sidorov, D., Teferle, N, Guerova, G., Egova, E., Vassileva, K., Milev, I., & Milev,
G. (2015). Sofia University GNSS Analysis Centre, FIG working week, 17–21/05/2015, Sofia,
Bulgaria. https://www.Figurenet/resources/proceedings/fig_proceedings/fig2015/papers/ts08g/
TS08G_simeonov_sidorov_et_al_7677.pdf
Simeonov, T., Sidorov, D., Teferle, F. N., Milev, G., & Guerova, G. (2016). Evaluation of IWV
from the numerical weather prediction WRF model with PPP GNSS processing for Bulgaria.
Atmospheric Measurement Techniques Discussions. https://doi.org/10.5194/amt-2016-152.
https://www.atmos-meas-tech-discuss.net/amt-2016-152/amt-2016-152.pdf.
Teunissen, P., Kleusberg, A., Bock, Y., Beutler, G., Weber, R., Langley, R. B., van der Marel, H.,
Blewitt, G., Gload, C., & Colombo, O. L. (1998). GPS for geodesy (2nd ed.). Berlin: Springer.
Tregoning, P., & Herring, T. A. (2006). Impact of a priori zenith hydrostatic delay errors on GPS
estimates of station heights and zenith total delays. Geophysical Research Letters, 33(L23303).
https://doi.org/10.1029/2006GL027706.
Trojáková, A. (2016). The NWP activities at CHMI, Joint 26th ALADIN Workshop & HIRLAM
All Staff Meeting 2016, Lisbon, Portugal, April 4–8.
Urquhart, L., Santos, M., Nievinski, F., & Boehm, J. (2011). Generation and Assessment of VMF1Type Grids using North-American Numerical Weather Models, XXV IUGG General Assembly,
Melbourne, Australia, June 28th – July 7th. Available at http://unb-vmf1.gge.unb.ca/Publica
tions.html
3 Advanced GNSS Processing Techniques (Working Group 1)
201
Václavovic, P., & Douša, J. (2015). Backward smoothing for precise GNSS applications. Advances
in Space Research, 56(8), 627–1634. https://doi.org/10.1016/j.asr.2015.07.020.
Václavovic, P., & Douša, J. (2016). G-Nut/Anubis – Open-source tool for multi-GNSS data
monitoring. In IAG Symposium. Berlin/Heidelberg: Springer. ISBN-13: 978-3-319-30895-1,
143, 775–782.
Václavovic, P., Douša, J., & Gyori, G. (2013). G-Nut software library – State of development and
first results. Acta Geodynamica et Geomaterialia, 10(4), 431–436. https://doi.org/10.13168/
AGG.2013.0042.
Václavovic, P., Douša, J., Elias, M., & Kostelecky, J. (2017). Using external tropospheric corrections to improve GNSS positioning of hot-air balloon. GPS Solutions, 21(4), 1479–1489. https://
doi.org/10.1007/s10291-017-0628-3.
Walpersdorf, A., Calais, E., Haase, J., Eymard, L., Desbois, M., & Vedel, H. (2001). Atmospheric
Gradients Estimated by GPS Compared to a High Resolution Numerical Weather Prediction
(NWP) Model. Physics and Chemistry of the Earth, 26, 147–152.
Warnant, R., Wautelet, G., Spits, J., & Lejeune, S. (2008) Characterization of the ionospheric
small-scale activity. Technical Report ESA, WP220, GALOCAD project.
Wautelet, G., Lejeune, S., Stankov, S., Brenot, H., & Warnant, R. (2008) Effects of small-scale
atmospheric activity on precise positioning. Technical Report ESA, WP230, GALOCAD
project.
Webb, F. H., & Zumberge, J. F. (1997). An Introduction to GIPSY/OASIS II. JPL D-11088.
Weber, G., Dettmering, D., Gebhard H., & Kalafus, R. (2005) Networked Transport of RTCM via
Internet Protocol (Ntrip) – IP Streaming for Real-Time GNSS Applications. In Proceedings
ION-GNSS 18th international technical meeting of the satellite division (pp. 2243–2247).
Weber, G., Mervart, L., Stürze, A., Rülke, A., & Stöcker, D. (2016). BKG Ntrip Client (BNC)
Version 2.12, Mitteilungen des Bundesamtes für Kartographie und Geodäsie, Band 49, ISBN
978-3-86482-083-0.
Wilgan, K., Hurter, F., Geiger, A., Rohm, W., & Bosy, J. (2017a). Tropospheric refractivity and
zenith path delays from least-squares collocation of meteorological and GNSS data. The Journal
of Geodesy, 91(2), 117–134. (licensed under CC BY 4.0, https://creativecommons.org/licenses/
by/4.0/).
Wilgan, K., Hadaś, T., Hordyniec, T., & Bosy, J. (2017b). Real-time PPP augmented with highresolution NWM model data. GPS Solutions, 23(3), 1341–1353. (licensed under CC BY 4.0,
https://creativecommons.org/licenses/by/4.0/).
Ye, S., Zhao, L., Song, J., Chen, D., & Jiang, W. (2018). Analysis of estimated satellite clock biases
and their effects on precise point positioning. GPS Solutions, 22, 16. https://doi.org/10.1007/
s10291-017-0680-z.
Zumberge, J. F., Heflin, M. B., Jefferson, D. C., Watkins, M. M., & Webb, F. H. (1997). Precise
point positioning for the efficient and robust analysis of GPS data from large networks. Journal
of Geophysical Research, 102, 5005–5017. https://doi.org/10.1029/96JB03860.
Zus, F., Dick, G., Heise, S., Douša, J., & Wickert, J. (2014). The rapid and precise computation of
GPS slant total delays and mapping factors utilizing a numerical weather model. Radio Science,
49(3), 207–216.
Zus, F., Dick, G., Douša, J., & Wickert, J. (2015a). Systematic errors of mapping functions which
are based on the VMF1 concept. GPS Solutions, 19, 277–286.
Zus, F., Dick, G., Heise, S., & Wickert, J. (2015b). A forward operator and its adjoint for GPS slant
total delays. Radio Science, 50, 393–405.
Related papers
Burned-out with burnout? Insights from historical analysis
Frontiers in Psychology
Alejo Carpentier. Supremul Magistrat sau Recursul la putere
Revista Transilvania, 2018
Performance evaluation of various hourly slope irradiation models using Mediterranean experimental data of Ajaccio
Energy Conversion and Management, 2006
Paolo Bordoni, Romeo Dell’Era, Simone Margnelli, “Le campane della Bregaglia”, Quaderni grigionitaliani 92/3, 2023, pp. 3-64.
Quaderni grigionitaliani, 2023
Violencia institucional contra las madres y la infancia Aplicación del falso síndrome de alienación parental en España
Ministerio de Igualdad, 2023
Harnessing AI for Transformative Business Intelligence Strategies
www.espjournals.org, 2023
Partecipazione e dialogo: il valore sociale ed educativo del museo
Heritage education: Cittadinanza e inclusione I, 2024
Yusuf Atılgan'ın Aylak Adam adlı romanında neden-sonuç ilişkisi
Uludağ Üniversitesi, 2017
Last phase of the Little Ice Age forced by volcanic eruptions
Nature Geoscience
Convergence and Optimization Results for a History-Dependent Variational Problem
Acta Applicandae Mathematicae, 2019
Marriage Forms in Vietnamese Fairy Stories Viewed from Ethnographic Perspectives
International Journal of Education and Social Science Research
La Incidencia Del Merchandising en El Contexto Mundial
Hitos de ciencias económico administrativas, 2019
Functional melatonin receptors in human prostate epithelial cells
Endocrinology, 1996
Hindered Ureas as Masked Isocyanates: Facile Carbamoylation of Nucleophiles under Neutral Conditions
Carolina Digital Repository (University of North Carolina at Chapel Hill), 2009