You are on page 1of 286

ELEMENTS OF GPS PRECISE POINT POSITIONING

By

Boonsap Witchayangkoon

B.Eng. (Honors) King Mongkut’s University of Technology Thonburi, Bangkok, 1992

M.S. University of Maine, 1997

A THESIS

Submitted in Partial Fulfillment of the


Requirements for the Degree of
Doctor of Philosophy
(in Spatial Information Science and Engineering)

The Graduate School


The University of Maine
December, 2000

Advisory Committee:

Dr. Alfred Leick, Professor of Spatial Information Science and Engineering, Advisor
Dr. Neil Comins, Professor of Physics and Astronomy
Dr. Ray Hintz, Associate Professor of Spatial Information Science and Engineering
Dr. Richard B. Langley, Professor of Geodesy and Geomatics Engineering,
University of New Brunswick
Dr. Charles Mundo, Adjunct Professor of Spatial Information Science and Engineering
Copyright 2000 Boonsap Witchayangkoon

ii
LIBRARY RIGHTS STATEMENT

In presenting this thesis in partial fulfillment of the requirements for an advanced

degree at the University of Maine, I agree that the Library shall make it freely available for

inspection. I further agree that permission for "fair use" copying of this thesis for scholarly

purposes may be granted by the Librarian. It is understood that any copying or publication of

this thesis for financial gain shall not be allowed without my written permission.

Signature: Boon Witchayangkoon

Date: August 25, 2000


ELEMENTS OF GPS PRECISE POINT POSITIONING

By Boonsap Witchayangkoon

Thesis Advisor: Dr. Alfred Leick

An Abstract of Thesis Presented


in Partial Fulfillment of the Requirements for the
Degree of Doctor of Philosophy
(in Spatial Information Science and Engineering)
December, 2000

The International GPS Service (IGS) now regularly makes accurate GPS ephemeris

and satellite clock information available over the Internet. The satellite coordinates are given

in the International Terrestrial Reference Frame (ITRF).

This thesis investigates Precise Point Positioning (PPP) using dual and single-

frequency pseudorange and carrier phase observations. Both the static and kinematic modes

are investigated. The static PPP solution examples use six-hour data sets from four stations.

The observations were made while Selective Availability (SA) was active and after it had been

discontinued. The static solutions agree to within 10 cm with published coordinates or with

solutions obtained from the Jet Propulsion Laboratory (JPL) PPP Internet service. The

kinematic solutions show a discrepancy of less than one meter, mostly around half a meter.

For observations with low multipath, the research shows that single-frequency

ionosphere-free PPP solutions are equivalent to the dual-frequency solutions. In case of

single-frequency observations the pseudorange dominates the solution.


Using a priori tropospheric information does not seem to improve dual-frequency PPP

solutions as compared to the case when the vertical tropospheric delay is estimated as part of

the Kalman filter solution. However, a priori tropospheric information seems to provide

benefits to single-frequency kinematic PPP. The Saastamoinen model is used when

computing the zenith tropospheric delay. In all cases, the Neill's mapping function is applied.

The studies show high correlation between receiver clock and the up coordinate. The

troposphere has a high negative correlation with receiver clock and the up coordinate.

However, the troposphere is more correlated with the receiver clock than the up component.

All solutions incorporate corrections for solid earth tides, relativity, and satellite

antenna phase center offsets. Corrections have not been applied for the phase wind-up angle.

The widelane and ionospheric functions are used to detect and fix cycle slips in a semi-

graphical manner. Since even a single cycle slip significantly falsifies PPP solutions, it is

suggested that between-satellite carrier phases be used as another way of detecting slips (now

since SA has been discontinued). The software consists mostly of highly modular Mathcad

functions that form an excellent base for continued research of PPP.


ACKNOWLEDGEMENTS

I would like to express my sincere gratitude towards my thesis advisor Dr. Alfred Leick for his

support and guidance, and the members of my thesis advisory committee: Dr. Richard

Langley, Dr. Ray Hintz, Dr. Chales Mundo, and Dr. Neil Comins, for their cheerful guidance

and cooperation during this thesis.

My thanks also go to numerous of people whom I may not name all, for their grateful

and joyful support, encouragement and helpfulness, especially Dr. Paulo Cesar Lima

Segantine and his family, Dr. Lee YoungChan and his family, Dr. James Zumberge, Dr. Peter

Kuntu Mensah and his family, Dr. Kate Beard, Dr. Ed Ferguson, Dr. Ramesh Gupta, Dr. Roop

Goyal and his family, Dr. Kridayuth Choompooming, Dr. Carol Bult, Dr. Hans-Georg

Scherneck, Zuheir Altamimi, Dennis Manning, Karen Kidder, Troy Jordan, Haci Mustafa

Palancioglu, Sharron J. Macklin, Dilnora Azimova, Jeanne Timmons, Karen and Robert

Liimakka, Mike Pearson, Teresa Cail, Puttipol Dumrongchai, Nakarin Satthamnuwong,

Balkaran Samaroo, Edward P. Wells, Brian J. Naberezny, Stephanie Sturtevant, Tom Noonan,

Carolyn Leick, Haci Mustafa Palancioglu, Samantha and Kurt Wurm, Pratya Levan, Wararat

Sophanowong, Ramaswamy Hariharan, Sawat Pararach, Taweesak Kijkanjanarat, Cheng Tee

Tang, Cynthia Henny, Jake Bogar, Tracey Nightingale, Young Su Kim, Saharat

Buddhawanna, Shinsuke Sasanawin, Auay Wanasen, Piriya Panwichai, Angsana Tokitkla,

Sunisa and Wattanachai Smittakorn, and Premwadee Furodchanakul.

I thank the Fogler Library for an excellent source of references for this research. I

also thank the Mathcad technical support for programming debug helps.

iii
Very special thanks are due to my parents for providing me excellent upbringing,

education, and guidance, which helped me reaching at this point in my life. Heartfelt thanks

go to my uncles, my aunts, my cousins, my sister, and my brothers, for their unconditional

support and understanding that always encouraged me to follow the path I have chosen.

Finally I would like to thank the Royal Thai Government and the Thammasat

University for financial assistance.

iv
TABLE OF CONTENTS

ACKNOWLEDGEMENTS ............................................................................................................III

LIST OF TABLES .........................................................................................................................XII

LIST OF FIGURES ..................................................................................................................... XIV

1 INTRODUCTION .................................................................................................................... 1

1.1 RESEARCH GOALS ............................................................................................................... 1

1.2 MOTIVATION....................................................................................................................... 3

1.3 PREVIOUS RELEVANT WORKS .............................................................................................. 4

1.4 APPROACH .......................................................................................................................... 9

1.5 THESIS ORGANIZATION........................................................................................................ 9

2 BACKGROUND..................................................................................................................... 11

2.1 THE GPS SYSTEM ............................................................................................................. 12

2.1.1 General Information ................................................................................................... 12

2.1.2 The Undifferenced Observation Equations .................................................................. 14

2.2 THE GLONASS SYSTEM ................................................................................................... 16

2.3 COMPONENTS OF PPP........................................................................................................ 18

2.3.1 Geophysical Models ................................................................................................... 19

2.3.2 Atmosphere................................................................................................................ 20

2.3.3 Reference Frames....................................................................................................... 23

2.3.4 IGS ............................................................................................................................ 23

2.3.5 Phase Wind-up Error .................................................................................................. 24

2.3.6 Receiver Antenna Phase Center Offset........................................................................ 27

2.3.7 Satellite Antenna Phase Center Offset and Satellite Orientation................................... 29

2.3.8 Satellite Clocks .......................................................................................................... 31

v
2.3.9 Group Delay Differential............................................................................................ 32

2.3.10 Relativity ................................................................................................................... 34

3 GEOPHYSICAL MODELS ................................ ................................ ................................ ... 36

3.1 DEFORMABLE E ARTH ........................................................................................................ 36

3.2 SOLID E ARTH TIDES .......................................................................................................... 37

3.2.1 Brief History .............................................................................................................. 37

3.2.2 The Tidal Potential..................................................................................................... 39

3.2.3 Solid Earth Tide Displacements .................................................................................. 42

3.3 OCEAN LOADING ............................................................................................................... 43

3.4 PLATE TECTONIC MOTION ................................................................................................. 46

3.5 ATMOSPHERIC TIDES ......................................................................................................... 49

4 INTERNATIONAL TERRESTRIAL REFERENCE FRAME (ITRF)................................ . 51

4.1 GENERAL STATEMENTS ON REFERENCE FRAMES ................................................................ 51

4.2 THE ITRF ......................................................................................................................... 53

4.3 TRANSFORMATION BETWEEN ITRFS .................................................................................. 53

4.4 ORIENTATION AND ORIGIN OF THE ITRF ............................................................................ 55

4.4.1 Orientation................................................................................................................. 55

4.4.2 Origin ........................................................................................................................ 56

4.5 THE DRAFT ITRF-2000 REFERENCE FRAME ....................................................................... 56

4.6 GPS WGS-84 ................................................................................................................... 57

4.7 AGREEMENT BETWEEN WGS-84 AND ITRF ....................................................................... 58

5 TROPOSPHERE AND IONOSPHERE................................ ................................ ................. 59

5.1 STANDARD ATMOSPHERE .................................................................................................. 60

5.2 TROPOSPHERE ................................................................................................................... 64

vi
5.2.1 Tropospheric Models.................................................................................................. 66

5.2.1.1 Hopfield Model ..................................................................................................... 67

5.2.1.2 Saastamoinen Model.............................................................................................. 69

5.2.2 Mapping Functions..................................................................................................... 72

5.2.2.1 Marini Mapping Function ...................................................................................... 73

5.2.2.2 Marini & Murray Mapping Function...................................................................... 73

5.2.2.3 Chao Mapping Function......................................................................................... 74

5.2.2.4 Lanyi Mapping Function (Lanyi) ........................................................................... 75

5.2.2.5 Davis Mapping Function (CfA-2.2)........................................................................ 76

5.2.2.6 Herring Mapping Function (MTT) ......................................................................... 78

5.2.2.7 Niell Mapping Function (NMF) ............................................................................. 78

5.3 IONOSPHERE...................................................................................................................... 81

5.3.1 Spatial and Temporal Variations................................................................................. 81

5.3.2 Ionospheric Range Delay............................................................................................ 85

5.3.3 Ionosphere Models ..................................................................................................... 87

5.3.4 Functions of Observables............................................................................................ 87

5.3.4.1 Dual-Frequency Ionosphere-free ............................................................................ 88

5.3.4.2 Dual-Frequency Ionosphere ................................................................................... 88

5.3.4.3 Single-Frequency Ionosphere-free Code and Phase................................................. 89

6 PRECISE IGS ORBIT AND SATELLITE CLOCK................................ ............................. 91

6.1 IGS ORBITAL ANALYSIS AND ITS PRODUCTS ...................................................................... 91

6.1.1 IGS Structure and Operation....................................................................................... 91

6.1.2 Products ..................................................................................................................... 92

6.2 THE SP3 EPHEMERIS ......................................................................................................... 98

6.2.1 The SP3 GPS Orbital Format and Data Accuracy........................................................ 98

6.2.2 Precise Satellite Clock Information............................................................................. 99

6.3 LAGRANGE INTERPOLATION .............................................................................................. 99

vii
7 MATHEMATICAL IMPLEMENTATIONS................................ ................................ ....... 101

7.1 DILUTION OF PRECISION .................................................................................................. 101

7.2 CYCLE SLIP DETECTION AND REMOVAL ........................................................................... 102

7.2.1 Multipath ................................................................................................................. 103

7.2.2 Widelane.................................................................................................................. 104

7.2.3 Cycle Slips ............................................................................................................... 105

7.2.3.1 Between Satellite Differences .............................................................................. 105

7.2.3.2 Undifferenced Observation Cycle Slip Detection and Fixing ................................ 110

7.2.3.1.1 Cycle Slip Detection in the Widelane Combination........................................ 110

7.2.3.1.2 Cycle Slip Detection in the Ionospheric Combination .................................... 111

7.2.3.1.3 Fixing Cycle Slip of Undifferenced Observation............................................ 112

7.3 KALMAN FILTER ............................................................................................................. 113

7.3.1 Extended Kalman Filter............................................................................................ 116

7.3.2 Discrete Gauss-Markov Process................................................................................ 118

7.3.3 PPP Implementation................................................................................................. 119

7.3.3.1 Partial Derivatives ............................................................................................... 119

7.3.3.2 Receiver Clock Estimation................................................................................... 120

7.3.3.3 Ambiguity Estimation.......................................................................................... 120

7.3.3.4 Observation Weighting Schemes.......................................................................... 121

7.3.4 Computational Flow and Software Components........................................................ 124

8 NUMERICAL STUDY AND RESULTS................................ ................................ .............. 127

8.1 DATA SETS ..................................................................................................................... 127

8.2 A PRIORI KALMAN FILTER SETTINGS ............................................................................... 129

8.3 ANALYSIS E XAMPLE........................................................................................................ 130

8.3.1 Widelane.................................................................................................................. 131

8.3.2 Ionospheric Carrier Phase......................................................................................... 132

8.3.3 Multipath ................................................................................................................. 132

viii
8.3.4 OMC for P1 ............................................................................................................. 133

8.3.5 OMC for P2 ............................................................................................................. 134

8.3.6 OMC Pseudorange Ionosphere-free .......................................................................... 135

8.3.7 Pseudorange OMC Difference Between Satellites ..................................................... 136

8.3.8 Pseudorange Ionospheric Variations ......................................................................... 137

8.3.9 Elevation Angle and Azimuth................................................................................... 138

8.3.10 Sky Plot ................................................................................................................... 139

8.3.11 North East & Up ...................................................................................................... 141

8.3.12 Receiver Clock Error Estimate.................................................................................. 142

8.3.13 Ionosphere-free Pseudorange OMC (Innovation) ...................................................... 142

8.3.14 Ionosphere-free Carrier Phase OMC (Innovation) ..................................................... 143

8.3.15 Reparameterized Ambiguity Estimates...................................................................... 144

8.3.16 Variance of Estimated Ambiguity ............................................................................. 145

8.3.17 Number of Satellites Used in Kalman Filter and DOPs.............................................. 146

8.3.18 Relativistic Correction.............................................................................................. 147

8.3.19 Approximated Tropospheric Effect Using Saastamoinen Model ................................ 148

8.3.20 Estimated Tropospheric Zenith Delay ....................................................................... 149

8.3.21 Correlation of the Estimated Parameters ................................................................... 149

8.4 EXPERIMENTS ................................................................................................................. 152

8.4.1 Discrepancies With Respect to Published Coordinates............................................... 152

8.4.2 Verification of Solid Earth Tides .............................................................................. 155

8.4.3 Estimated Troposphere vs. Saastamoinen Model....................................................... 156

8.4.4 Impact of Satellite Antenna Offset ............................................................................ 157

8.4.5 Impact of Relativity.................................................................................................. 158

8.4.6 Impact of Single Cycle Slip ...................................................................................... 159

8.4.7 Using A Priori Tropospheric Information.................................................................. 162

ix
8.4.8 Kinematic Positioning .............................................................................................. 164

8.4.8.1 Dual-Frequency Kinematic Positioning ................................................................ 164

8.4.8.2 Single-Frequency Kinematic Positioning.............................................................. 169

9 CONCLUSIONS AND RECOMMENDATIONS................................ ................................ 172

9.1 CONCLUSIONS ................................................................................................................. 172

9.1.1 Dual-Frequency Solutions ........................................................................................ 173

9.1.2 Impact of IGS Products and Service.......................................................................... 173

9.1.3 Single-Frequency PPP .............................................................................................. 174

9.1.4 Correlation Between the Up Coordinate, Zenith Tropospheric Delay, Receiver

Clock, and Ambiguities ............................................................................................ 174

9.1.5 A Priori Tropospheric Information............................................................................ 175

9.1.6 Kinematic PPP ......................................................................................................... 175

9.2 RECOMMENDATIONS ....................................................................................................... 176

9.2.1 Cycle Slip ................................................................................................................ 176

9.2.2 Phase Wind-up Error ................................................................................................ 176

9.2.3 Receiver Antenna Phase Center Offset...................................................................... 177

9.2.4 A Priori Tropospheric Information............................................................................ 177

9.2.5 GLONASS............................................................................................................... 177

REFERENCES ................................ ................................ ................................ ............................. 179

APPENDIX A: SP3 EPHEMERIS FORMAT................................ ................................ ............. 191

APPENDIX B: RELEVANT MATHCAD FUNCTIONS................................ ........................... 197

B.1 CYCLE SLIP DETECTION AND FIXING .................................................................................... 197

B.2 OMC COMPUTATIONS ......................................................................................................... 217

B.3 GRAPHING AND DATA EDITING ............................................................................................ 238

B.4 KALMAN ESTIMATION.......................................................................................................... 245

x
APPENDIX C: PSEUDORANGE PERFORMANCE USING PRECISE EPHEMERIS........... 263

APPENDIX D: HOURLY JUMPS IN RINEX FILES................................ ................................ 264

BIOGRAPHY OF THE AUTHOR................................ ................................ ............................... 265

xi
LIST OF TABLES

Table 2. 1 GPS satellite antenna phase center offset adopted by IGS

(Kouba and Springer, 1998) ..............................................................................29

Table 2. 2 Typical sample of TGD (extracted from the GPS broadcast navigation

message DOY2(2000))......................................................................................33

Table 3. 1 Relative contributions to tidal potential from various celestial bodies...................41

Table 3. 2 Sample of ocean loading file ................................................................................44

Table 3. 3 Cartesian rotation vector for each plate using the NNR-NUVEL1A

kinematic plate model (IERS Conventions (1996, p. 14). The units are

radians per million years.).................................................................................49

Table 4. 1 Transformation parameters from ITRF-94 to other ITRFs....................................55

Table 5. 1 Sea level standard values.....................................................................................60

Table 5. 2 The fundamental seven layers of the U.S. Standard Atmosphere 1976 .................62

Table 5. 3 Frequently used refractivity constants..................................................................66

Table 5. 4 Coefficients of the hydrostatic NMF mapping function (Niell, 1996) ...................79

Table 5. 5 Coefficients of the wet NMF mapping function (Niell, 1996) ..............................80

Table 6. 1. The current IGS structure/components................................................................93

Table 6. 2 Approximate Availability and Accuracy of the IGS Products...............................94

Table 6. 3 Comparisons of IGS Rapid and IGS Final combined EOP with the IERS

Bulletin A for 1997 (units: mas – milli-arc-sec.; ms - millisec.) (Kouba et

al., 1998) ..........................................................................................................97

Table 6. 4 Pre-computed denominators of li for the standardized fixed nodes....................100

xii
Table 8. 1 Station hardware ...............................................................................................128

Table 8. 2 Published coordinates of ARP ...........................................................................128

Table 8. 3 RINEX data sets................................................................................................129

Table 8. 4 Correlation coefficients of the estimated parameters [WES2,

DOY2(2000)] .................................................................................................150

Table 8. 5 The difference between the PPP solutions and the published coordinates

for station WES2 (cm) ....................................................................................153

Table 8. 6 The difference between the PPP solutions and the published coordinates

for station NJIT (cm).......................................................................................153

Table 8. 7 The difference between the PPP solutions and the published coordinates

for station USNO (cm) ....................................................................................154

Table 8. 8 The difference between the PPP solutions and the published coordinates

for station TSEA (cm).....................................................................................154

Table 8. 9 Averaged discrepancies before and after SA-off (cm) for dual-frequency

solutions .........................................................................................................154

Table 8. 10 The impact of satellite antenna offset on the station solutions (3D errors) ........158

Table 8. 11 Station coordinates offset due to ignoring relativistic effect .............................158

xiii
LIST OF FIGURES

Figure 2. 1 Instantaneous positioning error before and after SA (available from the US

Space Command (IGEB, 2000) ).......................................................................14

Figure 2. 2 Troposphere and other atmospheric layers...........................................................20

Figure 2. 3 Ionosphere regions..............................................................................................22

Figure 2. 4 Geometric effect on phase (Wu et al., 1993).......................................................25

Figure 2. 5 GPS satellite clock correction for PRN5 prior to and after SA

discontinuation..................................................................................................32

Figure 2. 5 Relativity corrections .........................................................................................35

Figure 2. 6 Plate tectonics....................................................................................................46

Figure 3. 1 Schematic of lunar tidal force (Vaní¹ek and Krakiwsky, 1982)...........................40

Figure 3. 2 Tidal Potential ...................................................................................................41

Figure 3. 3 Graphic representation of the M2 loading effect in vertical displacement

(Courtesy of Hans-Georg Scherneck, Onsala Space Observatory,

Chalmers University of Technology).................................................................45

Figure 3. 4 Major tectonic plates of the world .......................................................................48

Figure 5. 1 Hopfield single-layer polytropic model atmosphere............................................67

Figure 5. 2 Schematic of Saastamoinen tropospheric and stratospheric spherical

layered dry atmosphere .....................................................................................70

Figure 5. 4 Sunspot count 1700-1800 (top), 1800-1900 (middle), and 1900-2000

(bottom) (NOAA, 2000)....................................................................................82

Figure 5. 5 Sunspot number prediction for cycle 23 (NASA, 2000b) ....................................83

Figure 5. 6 Monthly mean sunspot numbers (NOAA, 2000).................................................84

Figure 5.7 shows GPS ionospheric range errors as functions of TECU and frequency. .........86

xiv
Figure 5. 7 GPS Ionospheric range errors as functions of TECU and frequency....................86

Figure 7. 1 Multipath on P1 and P2 for PRN29 [NJIT, DOY137(2000)].............................104

Figure 7. 2 Phase OMC between satellites without satellite clock correction applied.

L1 (top), L2 (middle), and ionosphere-free (bottom). The base satellite is

PRN25. [WES2, DOY138(2000)] ..................................................................107

Figure 7.3 Phase OMC between satellites with satellite clock correction applied.

L1 (top), L2 (middle), and ionosphere-free (bottom). The base satellite is

PRN25. [WES2, DOY138(2000)] ...................................................................108

Figure 7. 4 Phase OMC between satellite ionosphere-free PRN21-PRN25 without

(top) and with (middle) satellite clock correction applied, and satellite

clock correction difference (bottom). [WES2, DOY138(2000)] ......................109

Figure 7. 5 The Kalman filter computation recursive scheme .............................................115

Figure 7. 6 The EKF computation recursive scheme...........................................................117

Figure 7. 7 Potential weighting functions, comparison between cosecant and

cosecant-squared.............................................................................................122

Figure 7. 8 Potential step function weightings, comparison between equations 7. 33

(line) and 7. 34 (dotted line) ............................................................................124

Figure 7. 9 PPP Algorithm.................................................................................................126

Figure 8. 1 Widelane before and after cycle slip detection (top) and the

difference(bottom) for PRN5 [WES2, DOY2(2000)].......................................131

Figure 8. 2 Ionospheric carrier phase for PRN5 [WES2, DOY2(2000)] ..............................132

Figure 8. 3 Multipath on P1 and P2, for PRN5 [WES2, DOY2( 2000)] ..............................133

Figure 8. 4 OMC for P1 for PRN5 (top) and all satellites (bottom) [WES2,

DOY2(2000)] .................................................................................................134

xv
Figure 8. 5 OMC for P2 for PRN5 (top) and all satellites (bottom) [WES2,

DOY2(2000)] .................................................................................................135

Figure 8. 6 OMC for pseudorange ionosphere-free for PRN5 (top) and all satellites

(bottom) [WES2, DOY2(2000)] ......................................................................136

Figure 8. 7 OMC difference between satellites for P1 (top) and pseudorange

ionosphere-free (bottom). The base satellite is PRN5 [WES2,

DOY2(2000)] .................................................................................................137

Figure 8. 8 Pseudorange ionospheric variations PRN5 [WES2, DOY2(2000)]....................138

Figure 8. 9 Elevation and azimuth for PRN5 [WES2, DOY2(2000)] ..................................139

Figure 8. 10 Sky plot for PRN5 (top) and all satellites (bottom) [WES2,

DOY2(2000)] .................................................................................................140

Figure 8. 11 North (top), east (middle) and up (bottom) [WES2, DOY2(2000)]..................141

Figure 8. 12 Receiver clock estimate(top) and its variance (bottom) [WES2,

DOY2(2000)] .................................................................................................142

Figure 8. 13 Ionosphere-free pseudorange OMC (innovation) for PRN5 (top) and all

satellites (bottom) [WES2, DOY2(2000)]........................................................143

Figure 8. 14 Ionosphere-free carrier phase OMC (innovation) for PRN5 (top) and all

satellites (bottom) [WES2, DOY2(2000)]........................................................144

Figure 8. 15 Reparameterized ambiguity estimates PRN5 (top) and all satellites

(bottom) [WES2, DOY2(2000)] ......................................................................145

Figure 8. 16 Variance of estimated ambiguity PRN5 (top) and all satellites (bottom)

[WES2, DOY2(2000)] ....................................................................................146

Figure 8. 17 Number of SV used in the computation (top) and the respective DOPs

(bottom) [WES2, DOY2(2000)] ......................................................................147

Figure 8. 18 Relativistic correction for PRN5 [WES2, DOY2(2000)].................................148

xvi
Figure 8. 19 Approximated tropospheric error PRN5 (top) [WES2, DOY2(2000)] .............148

Figure 8. 20 Estimated zenith tropospheric error [WES2, DOY2(2000)] ............................149

Figure 8. 21 Correlation coefficients between the estimated parameters: clock-up-

troposphere (top), north-east-up(second from top), ambiguities (second

from bottom), ambiguity-north-east-up-troposphere (bottom) [WES2,

DOY2(2000)] .................................................................................................151

Figure 8. 22 Station solution (top) compared with the solid earth tides corrections

(bottom) [WES2, DOY2(2000)] ......................................................................155

Figure 8. 23 Comparison of tropospheric zenith delay between the estimated value

and the Saastamoinen model [WES2, DOY136(2000)]....................................156

Figure 8. 24 Adding a slip of one cycle to PRN29 at epoch 300 [USNO,

DOY138(2000)]..............................................................................................160

Figure 8. 25 Subtracting a slip of one cycle to PRN29 at epoch 300 [USNO,

DOY138(2000)]..............................................................................................161

Figure 8. 26 Solutions comparison between zenith troposphere estimated and

approximated [WES2, DOY136(top), 137(middle), and 138(bottom)

(2000)]............................................................................................................163

Figure 8. 27 Dual-frequency kinematic solutions comparison between troposphere

estimated and approximated (top) with the associated DOPs (middle) and

SV used (bottom) [WES2, DOY136(2000)] ....................................................165

Figure 8. 28 Horizontal offset comparison of kinematic solutions between zenith

troposphere estimated (top) and approximated (bottom) [WES2,

DOY136(2000)]..............................................................................................166

xvii
Figure 8. 29 Dual-frequency kinematic solutions comparison between troposphere

estimated and approximated (top) with the associated DOPs (middle) and

SV used (bottom) [USNO, DOY137(2000)] ....................................................167

Figure 8. 30 Horizontal offset comparison of kinematic solutions between zenith

troposphere estimated (top) and approximated (bottom) [USNO,

DOY137(2000)]..............................................................................................168

Figure 8. 31 Single-frequency kinematic solutions comparison between troposphere

estimated (top) and approximated (second from top) with SV used

(second from bottom) and DOPs (bottom) [WES2, DOY2(2000)] ...................170

Figure 8. 32 Horizontal offset comparison of kinematic solutions between zenith

troposphere estimated (top) and approximated (bottom) [WES2,

DOY2(2000)] .................................................................................................171

Figure C. 1 Pseudorange performance [WES2, DOY136 (top) and 137 (bottom)

(2000)]............................................................................................................263

Figure D. 1 Hourly jumps in pseudorange P1+P2 (top), but not carrier phase L1 + L2

(bottom) for PRN4 [GAIT, DOY2(2000)] .......................................................264

xviii
1 Introduction

1.1 Research Goals

The most popular GPS technique currently in use for accurate positioning is double

differencing. This technique is popular because common mode errors cancel for short

baselines or their impact is drastically reduced in the case of long baselines. This is

particularly true for satellite orbital errors and receiver clock errors. Another distinguishing

characteristic of double differencing is the relative ease of constraining double difference

ambiguities to integer values. Several ambiguity fixing strategies have become available that

typically rely on statistical tests and search strategies. The least squares ambiguity

decorrelation adjustment (LAMBDA) developed at the Delft University of Technology

(Teunissen, 1994) seems to enjoy an ever increasing popularity.

Single differencing has recently found a new value in connection with the processing

of GLONASS observations (Leick et al., 1998a and 1998b). Since GLONASS satellites

transmit generally at different frequencies, the receiver clock errors do not cancel in the

double difference carrier phase equations. Appropriate scaling prior to forming the double

differences results in cancellation of the receiver clock errors, but introduces a non-integer

double difference ambiguity. These difficulties can be avoided when processing single

differences and fixing double difference ambiguities implicitly.

1
Little emphasis has thus far been given to undifferenced processing of carrier phases,

i.e. using observation from one receiver only. The reason being is that common mode errors

do not cancel in this case. However, this situation is likely to change as highly accurate orbital

positions AND satellite clock data become available from the IGS (International GPS

Service). See Neilan et al. (1997) and Kouba et al. (1998) for a brief description of these

services. The availability of very accurate satellite clock data is of fundamental importance for

the undifferenced processing technique (the GPS satellite clock stability is intentionally

degraded when Selective Availability is active). The maturing of the IGS products makes it

opportune to subject precise point positioning (PPP) to a rigorous scientific study and analysis.

Per definition, PPP uses accurate orbital data AND accurate satellite clock data as provided

e.g. by the IGS, and DUAL-frequency pseudorange AND carrier phase observations collected

by the user, and uses the ionosphere-free function of the observables. PPP implies single

receiver positioning (as far as the user is concerned). The second receiver, needed in the

double difference approach, is essentially replaced by the collection of continuously operating

receivers around the world. The achievable accuracy of PPP approaches that for the

underlying network of receivers, i.e. we expect centimeter position accuracy.

The research addresses the following:

• Investigate the performance of PPP as a function of the accuracy of IGS orbits and

IGS satellite clock data. This research is expected to provide recommendations for

the design of future IGS products and services.

• Investigate the correlations between vertical coordinate, tropospheric zenith delay,

receiver time, and ambiguities. The research will shed light on the minimum

requirements to make the vertical parameter, the tropospheric zenith delay, the

receiver clock and the ambiguities independent (separable). The research will focus on

the accuracy of the vertical coordinate and on limitations resulting from biases that

2
might be unavoidable. The latter aspect is of particular interest for vertical

deformation applications.

• The performance of PPP will be investigated as a function of the length of

observations ranging from single epoch solutions (kinematic mode) to six hours.

• Investigate the possible contribution of connecting an atomic clock to the receiver.

Such a combination could be particularly useful for kinematic application.

• Clarify and investigate limitations and constraints of the PPP technique that may still

be caused by lingering Selective Availability.

• Investigate the utility of L1-only receivers for PPP (single-frequency PPP). In this

case the ionospheric correction requires special considerations.

• Study and document existing "geo" models that must be incorporated to make

centimeter positioning meaningful. Examples include solid earth tides, ocean loading,

plate tectonic motion, and atmospheric models.

• Develop Mathcad software to test components of PPP and support future research.

1.2 Motivation

Algorithmic developments as well as hardware developments have steadily improved since

about 1982 when civil applications of GPS positioning began. Highlights in these

developments include the introduction of dual-frequency receivers, narrow correlation

receivers, the antenna swap method, and OTF (ambiguity fixing On-the-Fly). The IGS has

quietly but effectively contributed to bringing GPS applications to a higher level of accuracy.

By virtue of having observation stations distributed worldwide and having organized

an efficient data management and analysis operation, the IGS now provides GPS orbital

accuracy at the 5 cm level (1σ). Equally important, it makes satellite clock information

3
available with correspondingly high accuracy. This is a major accomplishment that potentially

makes single receiver positioning competitive with relative positioning between two receivers.

Accomplishing one's high-accuracy "positioning mission" with just one receiver is probably as

revolutionary as the introduction of GPS itself. The major advantage of course is that the user

has to operate only one receiver at a time.

It is a characteristic of GPS that applications are too numerous to be listed. Many

applications are still emerging. For example, PPP is useful to researchers for studying diurnal

tidal motions of the solid earth center (Scherneck and Webb, 1998). Tregoning et al. (1998)

attempt to measure isostatic rebound in Antarctica using continuous remote GPS observations.

No doubt, these and other applications can benefit from PPP, primarily because single-

receiver operation and expected simplicity in processing. Since positioning with certain

accuracy implies time transfer capability with comparable accuracy, PPP is becoming a

preferred candidate for accurate time transfer. Already a major international campaign is in

preparation by the BIPM (Bureau International des Poids et Measures) and IGS (Ray, 1998).

Whereas research groups and research institutes are expected to pay increasing

attention to PPP and respective scientific applications, the proposed research specifically

keeps the needs of practicing surveyors in mind. Surveyors constitute a very large GPS user

group with specific needs. Not only will they appreciate operating only a single receiver, but

also the decreasing reliance on the CORS (Continuous Operating Reference Stations) and

other high accuracy reference networks.

1.3 Previous Relevant Works

Relevant work on PPP has been carried out primarily at JPL (Jet Propulsion Laboratory). In

fact the very term "Precise Point Positioning" seems to have its origin at JPL. There are two

4
publications from JPL researchers that are particularly relevant to PPP, indeed these papers

sparked the interested in the proposed research.

The theoretical foundation of PPP is documented in Zumberge et al. (1997a). They set

out to develop an efficient approach to make the "accuracy achieved at IGS for global

solutions" available to users. They recognized that various networks of stations have been

established around the world to satisfy the need for high accuracy positioning. An example is

the CORS network consisting of some 80 plus stations operated by the NGS (National

Geodetic Survey). These networks typically serve on the premise of relative positioning, i.e.

the user will position himself relative to the nearest CORS station. Sophisticated users might

reference to several CORS station. The JPL researchers state:

"To keep the computational burden associated with the analysis of such data

economically feasible, one approach is to first determine the precise GPS

satellite positions and clock corrections from a globally distributed network of

GPS receivers. Then, data from the local network are analyzed by estimating

receiver-specific parameters with receiver-specific data: satellite parameters

are held fixed at their values determined in the global solution. This 'precise

point positioning' allows analysis of data from hundreds or thousands of sites

every day with 40 Mflop computers, with results comparable in quality to

the simultaneous analysis of all data."

In order for users of PPP to achieve the highest possible accuracy, it is important that

the solution be "consistent." One not only must use their precise ephemeris and THEIR clock

data but also use the same "geo" models like earth tides, etc. to avoid a degradation of

achievable accuracy. Zumberge et al. (1997a) document the validity of their approach by

analyzing daily sets of carrier phase data achieving millimeter repeatability in the horizontal

components and centimeter precision in the vertical.

5
Even more astounding results are reported in Zumberge et al. (1997b). They

computed orbits and clock information solutions from the Flinn global network (this is a

subset of the IGS network). Many of the Flinn stations are equipped with a hydrogen maser or

a good quality rubidium or cesium clock. Thus a very stable time reference is available at the

receiver site. This is crucial when estimating high-rate satellite clock corrections (in

particular when SA is active). All Flinn network receivers record data at a 30 second interval.

GPS high-rate clocks are then estimated at the same respective 30-second epoch based on a

free-network solution for Flinn. They utilized such high rate clock information to analyze

carrier phase data from a single receiver for both static and kinematic mode. For the static

mode with 5 minutes of data, the 3-D positional accuracy was 0.44 cm, but with a daily

repeatability of 1.86 cm. The kinematic mode provided a 3-D positional accuracy of 3.4 cm.

This result is remarkable!

It seems clear that PPP constitutes a major step forward in the development of high

accuracy positioning, and that it is a complex technique. Zumberge and his colleagues are the

only researchers that have thus far reported PPP results of such a high accuracy. However,

there is no analysis of PPP reported in the literature and no explanation can be found why

other researchers seem to have failed to achieve comparable accuracy. The results reported

above were achieved with the software GIPSY/OASIS-II developed at JPL. Additional

information on this software is provided below. Currently, the Bernese software, but not

GAMIT, has PPP capability and the Bernese group is still experimenting with certain

components of PPP (Hugentobler, private communication). However, there is no evidence in

the literature.

Other researchers have previously reported results using PPP. However as we will see,

their results are at an order of magnitude worse than what has been achieved at JPL.

6
Hé roux et al. (1995) reported the accuracy of submeter for single point positioning

using pseudoranges in conjunction with the use of JPL's GIPSY/OASIS-II software and 30

second interval GPS orbits and clock corrections from the Geodetic Survey Division (GDS),

Natural Resources Canada (NRCan).

Lachapelle et al. (1996) applied kinematic single point positioning to the aircraft in

the post-processed mode. Single-frequency pseudoranges were used in the analysis and

precise ephemerides and clock corrections were also obtained from the NRCan. Using

GIPSY/OASIS-II software, the analytical results were then compared to DGPS and showed

consistent accuracy at 1 m (rms) in latitude and longitude and 2 m (rms) in height. They

concluded that the ionosphere degraded the accuracy, particularly in the height component

because a single frequency receiver was used.

Henriken et al. (1996) tested stand-alone positioning with single- and dual- frequency

pseudoranges. Using precise ephemerides and clock corrections from NRCan, the post-

processed analysis was conducted based on epoch-to-epoch solutions. The low passed

filtered Fast Fourier Transform (FFT) was then applied to remove high frequency receiver

noise. The results are accurate to 0.5-1.5 m horizontally and 1.5-3 m vertically depending

upon how one accounts for ionospheric corrections.

It must be noted that for the experiments listed above, the standard software was JPL's

GIPSY/OASIS-II. It should be noted that the Canadian experiments relied on pseudoranges

only. Pseudoranges are not as accurate as carrier phases; however carrier phases add

complexity because of additional ambiguity parameters and the possibility of cycle slips. The

Canadians apparently did not use a free-network solution, i.e. their results may have been

negatively affected by inaccuracies in the reference network. As IGS continuously improves

the solutions, the need for using a free-network solution as reference is expected to diminish.

7
In any case, PPP has been demonstrated by JPL at the couple of centimeter level and by

Canadian researchers at the sub-meter level. It is a technique to be reckoned with.

JPL's GIPSY/OASIS-II software has been developed typically in piecemeal fashion

over many years and by many people. Its primary purpose is satellite orbit and earth

orientation determination. It is the "working horse" for JPL's researchers. PPP capability has

been added apparently around 1994/95. The source code is Fortran that runs on UNIX. JPL

makes executables of GIPSY/OASIS-II available upon request. However, because GIPSY's

"shape and form" is that of an "internal research software", transporting it to other computers

and actually using it is no easy task. JPL has found it necessary to contract with Raytheon to

provide installation and consulting services to users (Zumberge, private communication). In

exceptional cases, the source code can be made available to researchers at universities or

government agencies. A formal contract between the University of California and the home

institute is required. The process can take up to several months (Zumberge, private

communication).

It appears that GIPSY/OASIS-II and the Bernese software are the only software

available that can deal with PPP. Even if the considerable effort is made to install

GIPSY/OASIS-II software package at the University of Maine, its value in terms of

supporting dissertation research is highly questionable. Dissertation research requires that the

source code is totally understood and modifications/additions to the program can be made

easily by the student. This does not seem to be the case with GIPSY/OASIS-II. It is,

therefore, unavoidable that a new, trimmed-down, dedicated PPP research software be

developed as part of this dissertation.

8
1.4 Approach

This research is involved with a study and utilization of existing geo-models to accomplish

centimeter positioning globally with GPS using IGS products. This study will further deal

with the current and planned IGS products to accomplish this goal. Respective software will

be developed and tested with real data. In order to facilitate the study, software components

developed by other researchers will be used. Examples are Neill’s tropospheric corrections

and the solid Earth tides software.

1.5 Thesis Organization

Chapter 2 provides background on the navigation satellite systems and the equations for the

basic undifferenced observables. Components of PPP are also summarized. The geophysical

models will be dealt with in detail in chapter 3. The treatment begins with the concept of

deformable earth followed by details on solid Earth tides, ocean loading, and plate tectonic

motion. Chapter 4 deals with the International Terrestrial Reference Frame (ITRF). This

encompasses the history of the development of geodetic reference frame, precise definition of

the current frame, and the effects of temporal variations. Chapter 5 discusses the tropospheric

and ionospheric effects on the GPS signals. The global mapping functions for tropospheric

corrections are given and the ionosphere-free functions of the observables are listed. Chapter

6 focuses on the precise IGS ephemeris and satellite clock information for GPS and gives the

Lagrange interpolation formulation. Chapter 7 addresses the mathematical implementation of

PPP. Two major sections deal with cycle slip detection for undifferenced observations and the

actual implementation of the Kalman filter. The numerical results are given in Chapter 8. This

is followed by conclusions and recommendations in Chapter 9. There are four appendices: the

9
SP3 ephemeris format, a collection of Mathcad functions developed for this study,

pseudorange solutions, and hourly jumps in RINEX files.

10
2 Background

The development of satellite-aided positioning or navigation is far from completed. While the

beginning of satellite positioning can be traced back to the days of the Sputnik satellite, the

Global Positioning System (GPS) has made satellite positioning available to a world-wide

community of users since the early eighties and is currently undergoing a modernization

phase. In the nineties the Russian GLObal NAvigation Satellite System (GLONASS)

commanded some attention. Combining the signals of both systems attracted the curiosity of

scientists to alleviate shortcomings of the individual systems. Since early 1998, the European

community has been planning to launch a similar satellite navigation system called Galileo.

This system is currently in the design phase (Hein, 2000).

In addition to paying attention to developing the space component of satellite systems

and refining positioning algorithms, a complete positioning infrastructure has been developed

consisting of world-wide and/or national reference networks, the IGS (International GPS

Service), and so on.

This chapter provides background on GPS and GLONASS and the various

components that are typical and essential for Precise Point Positioning.

11
2.1 The GPS System

2.1.1 General Information

GPS is a weather-independent 24-hour position and navigation system that is maintained and

operated by the U.S. Department of Defense (DoD). The first satellites of this system were

launched in 1978. The system achieved initial operability capability (IOC) in 1993 when the

orbital constellation reached 24 space vehicles orbiting at an altitude of about 20180 km and

55 degrees inclined orbital planes (Block I satellite at 64 degrees). The U.S. Air Force Space

Command (AFSC) announced full operational capability (FOC) in 1995 when the

constellation consisted of only Block II satellite.(USNO, 2000a). UTC(USNO) is the reference

for GPS time (not adjusted for leap seconds). The broadcast ephemeris refers to the WGS-84

geodetic reference frame.

The U.S. Coast Guard Navigation Center maintains a GPS home page (Navcen, 2000)

that is the best source for current information. This web address contains many downloadable

papers and reports, among others the ICD-GPS-200 which contains system specifications.

Since the first satellite was launched in 1978, there have been three generations of

GPS satellites, so-called Block I, Block II/IIA, and Block IIR. Block I and II/IIA were

manufactured by the Rockwell company, whereas Block IIR by Lockheed Martin. The next

GPS generation is called Block IIF, which the U.S. Air Force awarded a contract to Rockwell

in 1996. The Block IIF satellite, the latest generation of GPS managed by the NAVSTAR

GPS Joint Program Office at the Space and Missile System Center (SMC), has improvements

over previous blocks of GPS satellites including a design life of 12.7 years, a dramatic

increase in the growth space for additional payloads and missions, and provision for a new

high accuracy civilian signal. The contract calls for 33 satellites and is valued at about $1.3

billion. The first delivery of the Block IIF satellites is expected in 2005 at the earliest.

12
Some components of the GPS signal structure will be changed and supplemented as

part of on-going modernization efforts which aim to make the system perform even better and

more reliable for the two major, and yet distinct user communities, i.e. military and civilians.

Currently GPS transmits two carriers, L1 = 10.23 x 154 = 1575.42 MHz (wavelength

λ1 ≈ 19.0 cm) and L2 = 10.23 x 120 = 1227.6 MHz (wavelength λ2 ≈ 24.4 cm). The carriers

are modulated with a precision (P) code (L1 & L2), and a coarse acquisition code C/A (L1).

The P(Y)-code has been encrypted (Anti Spoofing AS) and is henceforth referred to as Y-

code. The chipping rates for the P-code and C/A-code are 10.23 MHz and 1.023 MHz

respectively. The C/A-code is normally available on L1 only, but could be activated on L2 by

the ground control. Upon completion of the modernization phase, the GPS satellites are

expected to transmit the C/A-code on L2 and have a new (third) civil signal, called L5 at

1176.45 MHz. Of course, L1 also carries the navigation message modulated at 50 bps.

The GPS system has two levels of service, Standard Positioning Service (SPS) and

Precise Positioning Service (PPS). SPS is a positioning and timing service continuously

available to (civilian) users worldwide with no charge. SPS refers to L1 C/A-code positioning

whereas PPS refers to P-code positioning. Until the recent discontinuation of selective

availability (SA), the broadcast ephemeris (epsilon) and the satellite clock (dither) were

intentionally falsified. As a result stated SPS positioning accuracy was 100 m horizontally and

156 m vertically, and the time transfer accuracy was 340 ns (95% probability) (USNO,

2000b). The White House decision resulted in the discontinuation (setting to zero) of SA at

4.00 UTC on May 2, 2000. Consequently, civilian C/A-code users have immediate access to

accuracy better than 20 meters (95%). Figure 2.1 shows the impact of discontinuing SA.

13
Figure 2. 1 Instantaneous positioning error before and after SA (available from the US

Space Command (IGEB, 2000) )

PPS, which is highly accurate and based on cryptographic changes of P(Y)-code, is

available to only authorized (military) users. By having a military P(Y)-code capable receiver,

PPS yields positioning accuracy (95% probability) of 22 m horizontally, 27.7 m vertically, and

200 ns for the time transfer accuracy to UTC (USNO, 2000b). Anti-spoofing (AS) has been

implemented to guard against unauthorized transmissions of satellite data that mimic the

actual satellite signals via the encryption of the P-code to form the Y-code.

2.1.2 The Undifferenced Observation Equations

Both the carrier phase and pseudorange observables are important to PPP. The relevant

expressions are in units of distance (e.g., Leick, 1995; Hofmann-Wellenhof, 1997):

c p
Φ ip,k = ρ kp − cdt k + cdt p + N + I kp,i ,Φ + Tkp
f i i ,k 2. 1
+ d k ,i ,Φ + d kp,i ,Φ + d ip,Φ + ε kp,i ,Φ

14
Pi ,pk = ρ kp − cdt k + cdt p + I kp,i , P + Tkp
2. 2
+ d k ,i , P + d kp,i , P + d ip, P + ε kp,i , P

where

i = subscript identifying L1 or L2

fi = frequency

k= receiver station identifier

p = satellite identifier

Φ k = measured carrier phase scaled to distance (meters)


p

Pkp = measured pseudorange

ρkp = geometric topocentric distance

p
N k = integer ambiguity

dtk = receiver clock error

dt p = satellite clock error

I kk,i,Φ , I kk,i,P = ionosphere for phase and pseudorange and frequency fi

Tkp = troposphere

dk ,i,Φ , dk ,i,P = receiver hardware delay for phase or range respectively

dkp,i,Φ , dkp,i,P = multipath for phase or range respectively

di,pΦ , di,P
p
= satellite hardware delay for phase or range respectively

ε kp,i ,Φ ,ε kp,i , P = random measurement noise for phase or range respectively

The tropospheric and ionospheric effects are discussed in a later chapter.

15
2.2 The GLONASS System

GLONASS is a Russia-based positioning system managed and maintained by the Russian

Space Forces. GLONASS is very similar to GPS. The full GLONASS constellation also calls

for 24 satellites. While FOC has briefly been achieved, there are currently only 10 GLONASS

healthy satellites (SFCSIC, 2000). The satellites are located in three orbital planes with 64.8

degrees inclination. At 19100 km orbital altitude their period is 11 hours and 15 minutes.

General information about GLONASS can be obtained from the Coordinated Scientific

Information Center, the Russian Space Forces web site (SFCSIC, 2000), and the ICD-

GLONASS (1998).

The GLONASS carriers L1 and L2 are also modulated with P-codes and C/A-codes.

In contrast to GPS, each L-band frequency is different for each GLONASS satellite as given

below (some GLONASS satellites on opposite sides of the orbit use the same frequency). For

satellite p being the GLONASS almanac number, and frequency channel number n, frequency

allocation for L1 and L2, respectively (ICD-GLONASS, 1998):

f1 p = (2848 + n) 0.5625 MHz , 2. 3

f 2p = (2848 + n) 0.4375 MHz . 2. 4

During 1998-2005, GLONASS applies frequency channels n = 0...12 without any restrictions,

but n = 0 and n = 13 are intended for technical purposes. Beyond 2005, GLONASS will use n

= (-7…+6), but n = +5 and +6 will be for technical purposes.

The GLONASS observation equations are similar to those of GPS, except that proper

identification of the carrier frequency now requires a subscript to identify the carrier and a

superscript to identify the satellite.

The reference frame for the broadcast ephemeris is PZ-90. Over the last 10 years

much effort has been made to relate the PZ-90 and WGS-84 coordinate systems. Typically

16
seven parameters were estimated to locate the origin (∆u, ∆v, ∆w), determine the orientation

(ε, β, ω) and scale (s). For example, Misra et. al. (1996) used

 u   ∆u   1 ω − β  x
 v  =  ∆v  + (1 + s ) − ω 1 ε   y  2. 5
    
 w ∆w  β −ε 1   z 

and obtained ∆u = 0, ∆v = 2.5 m., ∆w = 0, s = 0, ε = 0, β = 0, and ω = -1.9x10-6 radians.

GLONASS time is UTC(SU) and accounts for leap seconds according to the IERS

notification. Users are usually notified in advance of the planned corrections. During the leap

second correction, GLONASS time is also corrected by changing enumeration of second

pulses of all onboard cesium clocks. Difference between GLONASS time and UTC(SU) is

expected to be within 1 millisecond. Like GPS, GLONASS satellite clock correction is

included in the navigation message. Recent efforts to increase the GLONASS time

synchronization is reported in Mikhail (2001).

A systematic effort has been made through the International GLONASS EXperiment

(IGEX) campaign in 1998 to study and test the GLONASS system in general, derive the best

transformation parameters between PZ-90 and WGS-84, and to relate GLONASS time and

GPS time (Slater et al., 2000). Integrity monitoring of IGEX-98 data is reported by Jonkman

and Jong (2000a, 2000b, and 2000c).

After ending the experiment the participants agreed on a continuation. In late May,

2000, the campaign evolved into the International GLONASS Service Pilot Project (IGLOS-

PP), sponsored by the International GPS Service (IGS) (Slater, 2000). The pilot service will

operate for a period of up to four years, from 2000-2003. The combined GLONASS orbits

(IGEX-solution) as well as weekly summaries for the entire year of 1999 have been uploaded

17
to the CDDIS which is available on anonymous logon (CDDIS, 2000a). More information

about IGLOS-PP can be found at IGLOS-PP (2000).

The different frequencies of GLONASS cause extra complication when attempting to

fix double difference ambiguities (Leick and Mundo, 1997; Wang et al., 2000). These

difficulties are not relevant to PPP because one never attempts to fix undifferenced

ambiguities. The combined GPS/GLONASS PPP solution, however, would require separate

receiver clock parameters for GPS and GLONASS (even though the same receiver observes

the signals from both satellite systems). The research described in this thesis did not use

GLONASS observations.

2.3 Components of PPP

Unlike in relative positioning, common mode errors do not cancel in PPP. Station movements

that result from geophysical phenomena such as tectonic plate motion, Earth tides and ocean

loading enter the PPP solution in full, as do observation errors resulting from the troposphere

and ionosphere. Relevant satellite specific errors are satellite clocks, satellite antenna phase

center offset, group delay differential, relativity and satellite antenna phase wind-up error.

Receiver specific errors are receiver antenna phase center offset and receiver antenna phase

wind-up.

This section contains a brief description of these error sources and of other

components that are parts of PPP. Critically important components will be amplified in the

next chapters.

18
2.3.1 Geophysical Models

Tidal deformations occur in the solid Earth, in the oceans, and in the atmosphere, and all

three interact in different ways with the Earth’s rotation. In addition to causing the Earth's

rotation axis to precess and nutate in space, the gravitational lunar and solar attractions, in

combination with inertial force (centrifugal force) resulting from the orbital movement of the

Earth around the common gravity center of the Earth-Moon or Earth-Sun system, cause the

tides (Lambeck, 1980). The Earth deforms because it has a certain degree of elasticity. This

tidal component, called solid Earth tides, is accurately computable. The Earth tides are similar

to ocean tides, however the latter are strongly affected by coastal geography and ocean

topography. There are four measurable tidal constituents large enough for geodetic

consideration. These are the lunar diurnal, the lunar semidiurnal, the solar diurnal, and the

solar semidiurnal tides. Diurnal tides have a period of about 1-day (24 hours and 50 minutes)

whereas semidiurnal tides are about half a day (12 hours and 25 minutes).

Ocean loading, i.e. primarily vertical variation of the crust in primarily coastal areas,

is caused by sea level fluctuations due to the tides. In order to achieve truly centimeter-level

global geodesy, ocean loading must be included in the site positional analysis.

Tectonic plate motions have been the subject of intense geodetic research for several

decades using Satellite Laser Ranging (SLR), Very Long Baseline Interferometry (VLBI),

GPS, and other space geodetic systems. Normally, tectonic plate motion is detected by using

regional or global tectonic networks. Data tracking of observations over a long period of time

from one station to another yield vectors of plate tectonic displacements.

Currently, there are hundreds of IGS sites around the world recording daily data for

analyzing plate motion. In addition, there are 64 GPS sites in California which are being

applied to study earthquakes in the Los Angeles area. For each site, data is analyzed into time

19
series to see movement trends in terms of latitude, longitude and height. This shows the

evolution of positions and best-fitting velocities as can be examined over the Internet (JPL,

2000b).

2.3.2 Atmosphere

The tropospheric and ionospheric regions of the atmosphere affect the propagation of GPS

signals. The troposphere is the bottom portion of the Earth’s atmosphere, which is the layer

of weather on the Earth. The thickness varies between 8 - 16 km from pole to equator. The

tropospheric temperature normally varies inversely with height, approximately −6.5°C per km

(NOAA/NASA/USAF, 1976).

Figure 2. 2 Troposphere and other atmospheric layers

The troposphere normally has about 75% of the atmosphere’s mass and most of the

water vapor in the atmosphere. High water vapor concentration ranges from 4% in the

tropical regions (humidity about 60-80% or more), but diminishes to just trace amounts in the

polar areas. The average atmospheric pressure is 1.03 kg/cm2 holding nitrogen 78%, oxygen

21%, and other gases 1% (e.g., argon, hydrogen, ozone, and methane). There is a small

20
amount of carbon dioxide, but the concentration has doubled since 1900. Water cycling takes

place in the troposphere as the exchange and movement of water between the Earth's surface

and atmosphere. Solar energy causes water to evaporate, and wind circulates the moisture.

Air rises, then expands and cools down condensing water vapor and thus developing clouds.

Various types of precipitation happen depending upon size and temperature of water particles.

Consequently, the troposphere is changing according to temporal and seasonal variations.

Stratosphere layer and tropopause are above the troposphere. The stratosphere,

tropopause, and troposphere are considered the electrically neutral atmosphere, which is a

non-dispersive medium for radio waves at frequency less than about 20 MHz. The influence

of troposphere refraction on both the carrier phases and code modulation is identical.

However, a part of the signals’ energy is absorbed by non-ionized gases (e.g., carbon dioxide)

and water molecules. Consequently, these matters delay the signal up to 2.5 meters in the

zenith direction and 30 meters close to a horizon angle causing a longer signal travel time

from satellite to receiver as compared to vacuum. These delays vary with temperature,

pressure, and humidity as well as spatial and physical location of the receiver. Tropospheric

refraction cannot be eliminated with dual-frequency observation.

The ionosphere is a layer or layers of ionized air surrounding the Earth extending

from almost 80 km above the Earth’s surface to altitudes of 1000 km or more. The air is

extremely thin at these altitudes. When the atmospheric particles are ionized by radiation

(e.g., ultraviolet radiation and X-rays from the Sun), they tend to remain ionized due to few

collisions between free negatively charged electrons and positively charged atoms and

molecules called ions. These ions characterize the ionosphere. The free electrons affect the

propagation of radio waves, thus the GPS signals. Unlike the troposphere, the ionosphere is a

dispersive medium for radio waves, which means that the modulations on the carrier and

carrier phases are affected differently and this effect is a function of carrier frequency. The

21
impact decreases with the increased frequency. Normally, the radiated energy from a

transmitter gets through the ionosphere, in part absorbed by the ionized air, and in part

refracted or bent downward again towards the Earth’s surface. Further, carrier frequencies

below about 30 MHz are reflected by the ionosphere, thus only higher frequencies, such as

GPS signals, television and frequency-modulation (FM) radio, can normally penetrate the

ionosphere.

Figure 2. 3 Ionosphere regions

There are two distinct regions in the nomenclature of ionization as shown in Figure 2.3 . The

magnetosphere refers to the outermost region where the particle motion is controlled by the

geomagnetic field. The ionosphere can be divided into two main layers called the E layer

(sometimes called the Heaviside layer or Kennelly-Heaviside layer, from about 80 to 113 km.)

and the F layer (sometimes called the Appleton layer, which is above the E layer). The E

layer reflects low frequency radio waves while the F layer reflects higher-frequency radio

signals. The F layer is composed of two layers: the F1 and F2 layers, which start

approximately at 180 and 300 km above the Earth’s surface, respectively. The thickness of

the F layer changes at night, thus altering its reflecting characteristics (Jursa, 1985).

22
The dispersive characteristics of the ionosphere closely follows the 11-year cycle of

sunspots (see e.g. Knight et al., 1996; Klobuchar and Doherty, 1998; Kunches and Klobuchar,

1998).

2.3.3 Reference Frames

Many reference systems and frames have been introduced and made available to users.

Examples are WGS-84 and PZ-90 used by GPS and GLONASS respectively for the broadcast

ephemerides. The International Terrestrial Reference Frame (ITRF) has been established by

the International Earth Rotation Service (IERS). The ITRF frame is frequently updated

according to the new data obtained from various geodetic observation systems, thus producing

a time series of reference frames. The transformation from one reference frame to another is

generally accomplished with a seven-parameter transformation. This will be discussed in

Chapter 4. The IGS precise ephemeris is referenced to the ITRF.

2.3.4 IGS

The International GPS Service (IGS) was formally established by the International

Association of Geodesy (IAG) in 1993 and officially started its operations on January 1, 1994

after a successful pilot phase of more than one year (IGS 1997; Neilan et al., 1997). IGS is

composed of more than 200 globally distributed permanent GPS tracking sites, three Global

Data Centers, five Operational or Regional Data Centers, seven Analysis Centers, an Analysis

Center Coordinator, and a Central Bureau (IGS, 1998). Each IGS site of the global network

operates a dual-frequency GPS receiver that records measurements at 30-second intervals

(Zumberge et al., 1997b; and Neilan et al., 1997). The Jet Propulsion Laboratory (JPL) serves

as the Central Bureau and, since 1999, the Center of Orbit Determination in Europe (CODE)

23
serves as the Analysis Center Coordinator (Kouba et al., 1998). IGS is a member of the

Federation of Astronomical and Geophysical Data Analysis Services (FAGS) and it operates

in close cooperation with the International Earth Rotation Service (IERS). The IGS service

and its products are described in Chapter 6.

2.3.5 Phase Wind-up Error

Phase wind-up problem is associated with the antenna orientation, both at the satellite and at

the receiver. This is due to the electromagnetic nature of circularly polarized waves intrinsic

in the GPS signals. Ideally, at the receiver the measured angle of carrier phase equals the

geometric angle between the instantaneous electric field and a reference direction at the

receiving antenna. Thus, when the antenna orientation changes, so does the reference

direction, and subsequently the measured phase. Likewise, the change of satellite antenna

orientation changes the direction of the electric field at the transmitting antenna and, as a

result, the change at the receiving antenna, thus the measured phase.

Wu et al. (1993) derived the phase wind-up correction for a crossed dipole antenna,

but is applicable to more general cases. A crossed dipole antenna consists of two equal-gain

dipole elements perpendicular to each other. Let x̂ and ŷ be the unit vectors in the directions

of the two dipole elements at the receiver antenna (horizontal plane), see Figure 2.4.

Similarly, let x̂ ′ and ŷ ′ be the unit vectors in the directions of the two dipole elements at the

transmitting antenna. Symbol φ is a azimuth angle from the receiver antenna x-dipole

direction to the satellite and θ is a satellite zenith angle. Angles φ′ and θ ′ are at the satellite,

measured similar to that at the receiver.

24
At the receiving antenna, let the phase signal from the x-dipole be received 90° earlier

relative to that from the y-dipole element. The signals from both dipoles are added to form the

antenna output.

Figure 2. 4 Geometric effect on phase (Wu et al., 1993)

25
Undifferenced observation: The phase wind-up correction is given as

∆φ = φ + φ ′ + π 2. 6

π refers to the ground receiver pointing upward away from the center of the Earth.

Phase wind-up for single difference: Let the receiver apparatuses be set such that the x-axes

of the two antennas, k and m, point horizontally in the azimuthal direction along the baseline,

observing satellite p which is on the left of m when looking at k. The phase wide-up for single

difference is the difference of the phase wide-up of the two receivers:

∆φ SD = ∆φ mp − ∆φ kp 2. 7

Apply the associated phase wind-up corrections:

∆φ SD = φ mp − φ kp + (φ ' mp −φ ' mp )
= β1 − (π − α1 ) + γ 2. 8
= Ω1

in which α1, β1, and γ are the three inner spherical triangle angles formed by projecting k, m,

and p onto a unit sphere concentric with the earth. This result indicates that phase wind-up

correction for a single difference is equal to the inner area Ω1 of spherical triangle on a unit

sphere. If the satellite is on the other side of the baseline, the phase wind-up will have the

same magnitude, but opposite sign.

Phase wind-up for double difference: Similar to the single difference case, by adding

satellite q, the double difference phase-wind-up correction is

∆φ DD = ±Ω
2. 9
Ω = α + β + γ + δ − 2π

26
where α, β, δ, and γ are the inner angles of the spherical quadrilateral formed by the projection

of the two satellites and the two receivers onto the unit sphere.

Effect of carrier phase wrap-up induced by rotating GPS antennas has been studied by

Tetewsky and Mullen (1997).

2.3.6 Receiver Antenna Phase Center Offset

A recent study about GPS antenna phase center offset has been conducted at the National

Geodetic Survey (NGS) by Mader (1999). The effect occurs because a GPS range observation

is measured from a satellite transmitted signal to the electrical phase center of the receiving

antenna. The electrical phase center variation (PCV) is a function of a particular antenna's

phase pattern (Aloi, 1999).

GPS antenna phase center is neither a single well-defined physical point nor stable

spot, but rather varies with the changing direction of the incoming satellite signal. However,

practically, users assume that the received signal point stays constant over the observation

period, which is often referred to as the phase center of the antenna.

Mader (1999) experimented in a series of tests using baselines to study relative

antenna phase center position with respect to the reference antenna. But, absolute antenna

calibrations have not been clearly demonstrated. For very short baselines using identical

antennas at the opposite ends, the phase center variations should cancel out and no effect is

seen. On the other hand, when different antenna types are used and these variations are

disregarded, the baseline solution will be the weighted average of the individual phase centers

of the two antennas. The antenna phase center offset may reflect significant vertical

positioning accuracy of up to 10 cm and sub-centimeter in the horizontal.

27
Normally, PCV is a function of both elevation and azimuth (Wubbena et al., 1997;

Aloi, 1999). However, it is not easy to model PCV variations due to high temporal correlation

with signal reflection multipath and specific antenna. As a matter of simplicity by assuming

azimuthal symmetry, one simple model is rather to assume that the phase center varies as a

function of satellite elevation angle only.

If absolute antenna calibrations were known, it would be possible to include this

information with reference code and phase observations to position a physical point such as an

external antenna reference point (ARP):

fi p f
ϕ kp,i (t ) = ρ k (t ) + N kp,i + i Θ i ( E ) sin E + ε ϕ 2. 10a
c c

Pkp,i (t ) = ρ kp (t ) + Θ i ( E ) sin E + ε P 2. 10b

where Θ i (E ) is the calibrated vertical distance between phase center and ARP for Li (i = 1,

2). The symbol E denotes the satellite elevation angle. In equation (2.10) the clock,

tropospheric, ionospheric, and multipath terms are ignored for simplicity. The antenna phase

variations will effect the ionosphere-free observables as:

f12 f1 f 2
ϕ kp, IF (t ) = ϕ kp,1 (t ) − ϕ kp,2 (t )
f12 − f 22 f12 − f 22
f1 p f2 f f
= ρ k (t ) + 2 1 2 N kp,1 − 2 1 2 2 N kp, 2 2. 11
c f1 − f 2 f1 − f 2

[
+ f13Θ1 ( E ) − f1 f 22 Θ 2 ( E ) ]c ( fsin−Ef
2 2
1 2 )

28
f12 f 22
Pkp, IF (t ) = Pkp,1 (t ) − Pkp, 2 (t )
f12 − f 22 f12 − f 22
2. 12
[
= ρ kp (t ) + f12 Θ1 ( E ) − f 22 Θ 2 ( E ) ]f sin E
2
− f 22
1

2.3.7 Satellite Antenna Phase Center Offset and Satellite Orientation

Satellite antenna phase center offsets do not cancel for PPP and must be dealt with

accordingly. In double differencing, these offsets cancel. These offsets are given in the same

satellite-fixed coordinate system that is also used to express solar radiation pressure (Leick,

1995, p. 57). The origin of the coordinate system is at the satellite's center of mass, the k-axis

points toward the Earth center, the j-axis points along the solar panel axis, the i-axis completes

the right-handed coordinate system and lies in the Sun-satellite-Earth plane. This definition

breaks down when the Sun, the satellite and the Earth are colinear. In this case and when the

satellite is in the earth's shadow the satellite attitude becomes unstable and complicated to

model (Bar-Sever, 1996).

Starting on 1998-Nov-29 (GPS Week 986, day 0) the IGS products incorporated the

antenna phase center offsets given in Table 2.1.

Table 2. 1 GPS satellite antenna phase center offset adopted by IGS


(Kouba and Springer, 1998)

Block II/IIA: (i, j, k) = (0.279m, 0.000m, 1.023m)

Block IIR : (i, j, k) = (0.000m, 0.000m, 0.000m)

The satellite ephemeris refers to the center of mass of the satellite. It can readily be

envisioned that the k-offset will be absorbed by the receiver clock estimate if not corrected

29
properly. GPS 43 (PRN 13) was the only Block IIR satellite available at the time this research

was conducted.

Let XSat and XSun be the GPS satellite and the Sun coordinates in the ECEF system. XSat

is obtained from the SP3 satellite ephemeris and XSun (or Xmoon) is calculated from the planetary

ephemeris. Unit vector e at the satellite and pointing towards the Sun is,

v X − X Sat
e = Sun 2. 13
X Sun − X Sat

Unit vector at the satellite center of mass and pointing to the Earth's center is,

v − X Sat
k= 2. 14
X Sat

The unit vector along the solar panel axis is,

v v v
j = k ×e 2. 15

The direction that completes the satellite-fixed right handed coordinates system is,

v v v
i = j×k 2. 16

If O denotes the antenna phase offset expressed in the satellite fixed (i, j, k) coordinate system

and given in Table 2.1 then,

∆X Sat = R −1 O 2. 17

is the offset expressed in the ECEF coordinates system where the rotation matrix R is:

v
i T 
v 
R =  jT  2. 18
v
k T 
 

30
The satellite phase center antenna in ECEF is

X SV = X Sat + ∆X Sat 2. 19

2.3.8 Satellite Clocks

Taking the speed of light as approximately 3 x 108 m/s, a satellite clock error of 1 µs causes an

error in the computed topocentric distance of 300 m. Accurate knowledge of the satellite

clock errors is of central importance to PPP. Without knowing the satellite clock error there

would be no PPP technique because the respective solution would be in the same "class" as

the standard navigation solution (which only corrects the satellite clock errors as provided in

the broadcast message). Of course, one major advantage of single and double differencing is

the elimination of the satellite clock error.

Each of the Block II/IIA satellites carries two cesium (CS) and two rubidium (RB)

atomic clocks. For Block II/IIA, CS clocks are considered the best satellite clocks (USNO,

2000a). One of the atomic clocks defines space vehicle time (others operate as spare). The

satellite clock errors are estimated by the IGS in connection with their satellite ephemeris

production using a global data set. The SP3 ephemeris files contain a respective column for

the satellite clock. Figure 2.5 shows a typical example of a satellite clock variation computed

by JPL under the influence of SA and after discontinuation of SA.

31
Figure 2. 5 GPS satellite clock correction for PRN5 prior to and after SA

discontinuation

2.3.9 Group Delay Differential

Group delay differential (TGD) is the L1-L2 instrumental bias that differs from satellite to

satellite. The L1-L2 correction is given by bits 17 through 24 of word seven of the navigation

message. TGD is pre-calculated by the Control Center based on measurements made by the SV

contractor during factory testing (ICD-GPS-200C). The value of TGD equals 1 /(1 − γ ) times

the delay differential, i.e.:

t Pp1 − t Pp2
TGD = 2. 20
1−γ

32
where γ = ( f1 f 2 ) 2 and t p is the time when the signal for each frequency is transmitted.

Single frequency users must correct the satellite clock (as computed from the

polynomial coefficient given in broadcast message) as follows: (ICD-GPS-200C):

dt Lp1 = dt p − TGD 2. 21a

dt Lp2 = dt p − γ TGD 2. 21b

Single frequency users that process pseudoranges and carrier phases for PPP do not have to

correct for TGD because they are absorbed by the estimated ambiguities. In case of dual-

frequency observation substituting (2.21) into the dual-frequency ionosphere-free function

(5.40) cancels TGD .

Table 2. 2 Typical sample of TGD (extracted from the GPS broadcast navigation message
DOY2(2000))

PRN TGD (ns)


1 -3.259629011154
2 -1.396983861923
4 -6.053596735001
5 -4.190951585770
6 -5.122274160385
7 -1.862645149231
8 -4.190951585770
9 -5.587935447693
10 -1.862645149231
13 -1.210719347000
16 -9.313225746155
17 -1.862645149231
18 -5.122274160385
24 -9.313225746155
26 -6.519258022308
27 -4.190951585770
30 -7.916241884232

Currently JPL provides updated estimates of TGD to the US Air Force Second Space

Operations Squadron (2SOPS) every quarter and also monitors the values daily to identify any

33
abrupt changes in the TGD values due to configuration changes on the satellites. The first

complete set of biases were uploaded on 29 April 1999 (Wilson, 1999). Example of TGD

values given in the broadcast ephemeris for DOY2, 2000, is shown in Table 2.2.

Having the correct TGD values provides three important benefits (Wilson, 1999):

1) single-frequency users who are not subject to SA (which has been discontinued) gain

higher positioning accuracy because the satellite clock error can be computed more

accurately.

2) For WADGPS, it provides more consistent use of fast clock corrections.

3) For the ionospheric community, when the L1-L2 bias for the receiver is known or

estimated, it increases the capability to extract the absolute TEC from dual-frequency

observations.

2.3.10 Relativity

Because of GPS orbital eccentricity, it is necessary to take into account the small relativistic

clock correction as suggested in the ICD-GPS-200C:

2
dt rel = X ⋅ X& 2. 22
c2

where X and X& are position and velocity of a GPS satellite.

Relativistic correction changes from satellite to satellite and from epoch to epoch are

seen in Figure 2.5. Unlike in differential GPS, the relativistic corrections must be applied in

PPP.

34
Figure 2. 5 Relativity corrections

35
3 Geophysical Models

3.1 Deformable Earth

The Earth is primarily composed of three basic components: solid (e.g., rock), liquid (e.g.,

ocean), and atmosphere. These components make the Earth far from absolutely rigid.

Therefore, in order to study the Earth’s deformation, the more realistic Earth’s model should

be somewhere in between being rigid (all considered solid) and being liquid. The Earth must

be seen as a deformable body over a wide range of time scales in response to changing surface

loads in the atmosphere, ocean, and hydrosphere (Lambeck, 1989). In addition, we have the

knowledge about the Earth’s motion that the Earth revolves around the Sun, together with

other planets, and at the same time the Earth also rotates or spins around its instantaneous axis

of rotation. The Earth’s motion, combined with solar and lunar attraction forces, causes Earth

tides, which are of interest regarding their effects on geocentric coordinates. In general, the

Earth’s temporal deformations occur locally as well as globally. Tidal deformations take

place in the solid Earth, in the oceans, and in the atmosphere.

There are two potentials relevant for each point of the Earth’s surface. First, the

gravity potential results from the Newtonian attraction from the whole mass of the Earth.

Second, the Earth's rotation causes the centrifugal potential. The forces corresponding to the

36
difference in the potential cause the tides. The oceanic tides are characterized by the fact that

the sea surface steadily adjusts itself to the potential surface.

The Earth’s physical properties are rather quite complex (elasticity, viscosity, and

plasticity) and cause the Earth to react to forces in a complex manner. In terms of temporal

variations the shape of the Earth and consequently the positions of the points on it can be

classified into three categories according to Vaní¹ek and Krakiwsky (1982):

• Secular – linear, slow, creeping

• Periodic – with period ranging from fractions of a second up to tens of years

• Episodic – suddenly accelerating and decelerating

In the following sections, we are concerned with Earth tides, ocean loading, and plate

tectonic motion.

3.2 Solid Earth Tides

3.2.1 Brief History

The following succinct history about the solid Earth tides is extracted from the first chapter of

Melchor (1978 and 1983).

"In 1824, the mathematician Abel was the first who pointed out that the

direction of the vertical does not stay constant but changes according to the

influence of the attraction forces from the Sun and the Moon. In 1844, C.A.

Peters published the first calculation of this effect. Around 1876, Lord Kelvin

drew attention to the deformable effects of the Earth itself indicating that it

was no longer acceptable to consider the Earth as being completely rigid.

37
Kelvin then demonstrated that the amplitudes observed at the Earth’s surface,

for each phenomenon derived from the tidal potential (i.e., oceanic tides,

deviations of the vertical, variations of the gravity force) would be affected by

the deformation of the surface on which our measurements are made.

In the early nineteenth century, the thought of a not entirely rigid Earth, but

deformable as a result of the tides, had therefore begun to be accepted. In the

meantime, some astronomers doubted the periodic variations in the latitudes

and studied the oscillation of the direction of the vertical and the local

deformations of the Earth crust. The horizontal pendulum, invented in 1832

by Hengler and Zöllner, was the first basic instrument in the study of Earth

tides and Seismology. Due to imperfections of the suspension wire available

at that time, von Rebeur Paschwitz conceived a suspension on metallic points.

This became the first instrument to record deviations of the vertical caused by

the Earth tides. In 1890, research institutes of Potsdam, Strasbourg, and

Teneriffe equipped with Paschwitz’s instruments demonstrated that there is an

existence of periodic oscillations of the vertical. Concurrently, Küstner and

Marcuse were experimentally showing the actual existence of the periodic

displacements of the instantaneous rotation axis of the Earth. Surprisingly,

the two phenomena of the variations of latitude and Earth tides governed by

the same theory were discovered at the same epoch in the same country,

Germany.

The simplest method that provides a clear demonstration that bodily tides do

exist is rested on a very simple logic. When observing ocean tides relative to

the marks fixed on the crust accompanied with the tide gauge, these marks

would be perfectly fixed if the globe was perfectly rigid, and the observed

38
amplitude of the ocean tide would then be equal to that calculated. On the

other hand, if the solid part is deformed, then the measured amplitude will be

equal to the difference between oceanic and Earth tides. G. Darwin

implemented this procedure in observing long period oceanic tides (monthly

and semi-monthly lunar tides)."

3.2.2 The Tidal Potential

The Earth tides occur according to the variations of the gravitational force, or the so-called

tidal force, exerted by celestial bodies such as the Moon. For any point on or within the

Earth’s surface, the gravitational force exerted by a celestial body is a sum of two components,

see Figure 3.1. The first component is the force that governs the Earth’s motion as a whole.

This equals the gravitational force acting at the Earth’s center of gravity. The second

component is the tidal force that equals the remainder of the force acting at the considered

point. Interestingly, the tidal force at a far point (e.g., point D) acts in the outward direction to

the celestial body. The reason is that the Earth is accelerating towards the attracting body at

the same rate as its center of mass M, but the near side (e.g., point A) is accelerating more and,

on the contrary, the far side (e.g. point D) is accelerating less than the center of mass.

Viewing all as a whole, the tidal force attempts to deform the equipotential surface of the

Earth’s gravity field causing its shape to prolate in the celestial direction. In other words, the

shapes are likely to elongate in the direction of the resultant force exerted by the configuration

of the celestial bodies.

39
Figure 3. 1 Schematic of lunar tidal force (Vaní¹ek and Krakiwsky, 1982)

Beginning with Newton's law of gravitation, the tidal potential can readily be derived.

Vaní¹ek and Krakiwsky (1982) give the expression:

n
G m ∞  r 
Wt ( P ) =
m
m ∑  d Pm  Pn (cos Z P ) 3. 1
dM n= 2  M 

The symbols have the following meaning:

m the mass of the Moon, 7.38 x 1022 kg

G the universal constant of gravitation, 6.673 x 10-11 m3/(kg s2).

d Mm average distance between center of the Earth and center of the Moon

ZP the zenith angle of the Moon at P

Pn associated with a series of the Lagendre's functions

rP the distance from center of mass of the Earth to point P

40
Figure 3. 2 Tidal Potential

A similar equation can be obtained for the solar tidal potential by simply replacing

the notation of Moon (m) with the Sun (s). Typically, the solar potential is about 46% of the

lunar potential. Other celestial bodies contribute much less as shown in Table 3.1.

Table 3. 1 Relative contributions to tidal potential from various celestial bodies


(Vaní¹ek and Krakiwsky, 1982)

Celestial Bodies Tidal Potential


Moon 1.0
Sun 0.4618
Venus 0.000054
Jupiter 0.0000059
Mars 0.0000010

As it should be expected at any point in and on the Earth, the luni-solar potential

varies temporally. This is primarily due to the temporal changes in geocentric distances

d Mm , d Ms and zenith distances Z m , Z s . The largest amplitudes of these periodic variations

are semidiurnal and diurnal. Diurnal period band is caused by lunar and solar motion.

41
Whether the celestial body is overhead or under the observer, the semidiurnal results in an

identical tidal potential. Lunar semidiurnal, which has the period of half a lunar day, is the

major contribution to the tidal potential.

3.2.3 Solid Earth Tide Displacements

Taking the mathematical gradient of the tidal potential (3.1), the tidal force components in the

geocentric coordinate system are obtained. These force components are relatively easy to

compute because only the well known expressions for the celestial motion of the sun and the

moon are needed. To convert the force components to actual displacements requires

knowledge of the so-called Love and Shida numbers. These numbers are "conversion factors"

that reflect the non-rigidy of the Earth, or reflect the yield of the actual earth to the tidal

forces. Because of the complexity of the Earth's deformation property these numbers have

been determined experimentally and are continuously getting refined (IERS Conventions,

1996).

Observed positions on the surface of the solid Earth must be corrected for solid Earth

tide displacement in order to obtain coordinates in the time-invariant ITRS reference frame.

For example,

 x0   x(t )   ∆x (t ) 
 y  =  y (t ) − ∆y (t ) , 3. 2
 0    
 z 0   z (t )   ∆z (t ) 

where x(t ), y (t ), and z (t ) are coordinates of an observed position at time t. The solid Earth

tides corrections ∆x(t ), ∆y (t ), and ∆z (t ) are obtained by using JPL's Development

Ephemeris DE403 (planetary) and Lunar Ephemeris LE403 (JPL, 2000a), and a Fortran

program downloaded from IERS96 (2000).

42
3.3 Ocean Loading

The Ocean loading tide is the deformation of the sea floor and adjacent land responding to the

redistribution of seawater which takes place during the ocean tide (Zlotnicki, 1996). The pure

ocean tide can primarily be measured by using tide gauges, whereas altimeters measure the

sum of ocean, loading and Earth body tides (Zlotnicki, 1996). The pure ocean tide can be

directly observed at the beach from rising and falling with respect to a benchmark. The tide

gauges installed at the coastlines measure and record these transitions. One can also put a

pressure gauge at the ocean floor to detect the dynamics of ocean tide. The sum of pure ocean

tide, loading tide, and Earth tide is called geocentric tide, which can be sensed from space

using an altimeter. Elastic ocean tide is the sum of the ocean and ocean loading tide

(Zlotnicki, 1996).

A site displacement component ∆c (radial, west, south) at a particular site at time t can

be written as

∆c = ∑ f j Acj cos(ω j t + χ j + u j − Φ cj ) , 3. 3
j

where

ψ colatitude

χj mean longitude of Sun, Moon, lunar perigee

fj , uj functions of longitude lunar node

Acj , Φcj site-specific elements that reflect the coastal geography, the

elastic and density structure of an earth model.

43
The subscript j runs over the major lunar and solar tides. They are: M2 (principle

semidiurnal), S2, N2, K2 (semidiurnal), K1, O1, P1, Q1, (diurnal), and Mf, Mm, and Ssa (long-

period).

Table 3.2 shows a sample of ocean loading file for station WES2, Westford,

Massachusetts (OSO, 2000). Figure 3.3 shows a snapshot of ocean loading along the East

coast.

Table 3. 2 Sample of ocean loading file

$$ COLUMN ORDER: M2 S2 N2 K2 K1 O1 P1 Q1 MF MM SSA


$$ "PTM" = COMBINED SOLUTION:
$$ From RRAY: M2 S2 N2 K2 K1 O1 P1 Q1
$$ From SCHW: MF MM SSA
$$ ROW ORDER:
$$ AMPLITUDES (m)
$$ RADIAL
$$ TANGENTL EW
$$ TANGENTL NS
$$ PHASES (deg)
$$ RADIAL
$$ TANGENTL EW
$$ TANGENTL NS

WES2 WES2
$$ GOT99.2_R.Ray_CC_PP_PTME ID: Feb 3, 2000 16:12 PTM
$$ Computed by H.G.Scherneck on gere.oso.chalmers.se, 2000
$$ 40440S020 P WES2, IGS, GPS RADI TANG lon/lat: 288.5062 42.6129
.00716 .00193 .00191 .00067 .00422 .00281 .00139 .00048 .00043 .00015 .00060
.00353 .00065 .00089 .00019 .00047 .00022 .00015 .00003 .00011 .00004 .00020
.00188 .00041 .00039 .00015 .00033 .00034 .00011 .00011 .00003 .00003 .00013
-171.3 -163.1 173.1 -162.9 -9.4 -3.2 -8.9 3.6 12.8 61.4 -86.2
-129.2 -127.3 -156.2 -134.0 -21.4 15.7 -19.6 66.7 -3.2 170.7 -95.2
-22.5 9.0 -37.1 4.3 173.3 -155.3 176.5 -171.4 -55.7 -93.0 26.9

Ocean tide loading is the largest perturbation in the solid Earth tide predictions. Both

amplitude and phase of ocean loading effects are heavily station and frequency dependent,

normally having magnitude of centimeters, and where the vertical displacement is

approximately three times larger than the horizontal components.

The conventional IERS models to compute ocean loading displacements (IERS

Conventions, 1996) do not include the motion of the origin of the coordinate system (motions

44
of the center of mass), but contain only the displacements due to deformation with respect to

the center of gravity of the solid earth (Scherneck, 1998a).

Figure 3. 3 Graphic representation of the M2 loading effect in vertical displacement

(Courtesy of Hans-Georg Scherneck, Onsala Space Observatory, Chalmers

University of Technology)

45
3.4 Plate Tectonic Motion

The Earth’s surface layer thickness ranging from approximately 40 to 90 kilometers is

assumed to be composed of a set of large and small plates all together consisted the rigid

lithosphere (the outer part of solid Earth) having average density 2.67 x 103 kg/m3. The

lithosphere lies on and slides over an underlying, weaker layer of partially molten rock

resulting from heat and pressure, called the asthenosphere, with density 3.27 x 103 kg/m3. The

lithosphere plate movements across the surface layer of the Earth are driven by stress forces

and interact along the plate boundaries producing divergence, convergence, or slippage of

plate boundaries.

Figure 2. 6 Plate tectonics

Historically speaking, around the eighteenth century, Leonhard Euler, Swiss

mathematician, described Earth’s surface plate movement by using the spherical geometry

theorem which describes such movement as a rotation around the pole. About 1908 to 1917,

the German geologist Alfred Wegener proposed the theories of continental drift. He along

with others also recognized that continental plates rupture, drift apart, and eventually collide

46
with each other. Plate tectonic has had a pervasive impact on Earth sciences since around

1967-68 when geodetic space techniques became available.

Plate tectonic theory is associated with lithosphere which is divided into a small

number of plates that float on or move independently over the Earth’s mantle. The sizable and

sudden plate’s movement causes Earthquake due to its stresses and/or volcanic activity. The

nature of plate tectonic activity during most of the Earth history is still ambiguous.

The plate tectonics hypothesis has developed to synthesize Earth’s dynamic behavior,

thus simplifying plate tectonics concepts. Geodesy has provided an important role for plate

tectonics study with high temporal resolution of the plate movements, particular from space

state-of-the-art technologies such as GPS and VLBI. Lambeck (1989) has given axioms of the

plate tectonics hypothesis as follows:

§ The plate tectonic motions are uniform on time scale of a million years or longer, but

this may be an artifact of the resolution of the geological observations. This agrees

with many recent geodetic observations that the present-day plate motions are very

similar to average motions for the past few million years.

§ All inter-plate motion occurs on the plate boundaries. Geodetic technologies are used

to observe how the motions between adjacent plates are absorbed, which relates stress

and strain fields across the plate boundaries.

§ Considering the points away from their boundaries, the plates function essentially as

rigid bodies, moving relative to each other without experiencing distortion. This

implies that either the deformations are small compared with the motions at the plate

boundaries, or these internal distortions are small when averaged over periods of

millions of years.

47
These key hypothesis are important in understanding plate tectonics process involved

in geodesy. Other assumptions can also be made to facilitate the plate tectonics study. The

modern geodetic observations are able to answer to what extent these hypothesis are valid.

Plate Motion Model: The IERS96 recommends the NNR-NUVEL1A model for the plate

motions given by DeMets et al. (1994).

Figure 3. 4 Major tectonic plates of the world

(Courtesy of the Hawaii Natural History Association) (HNHA, 2000)

Figure 3.4 shows map of the tectonic plates. Cartesian rotation vector for each plate of the

NNR-NUVEL1A kinematic plate model is given in Table 3.3.

The actual transformation of Cartesian coordinates (X0, Y0, Z0) and (X, Y, Z) of the

epochs t0 and t is given by

X = X 0 + 10 −6 [Ω Y Z 0 − Ω Z Y0 ](t − t 0 )
Y = Y0 + 10 −6 [Ω Z X 0 − Ω X Z 0 ](t − t 0 ) . 3. 4
Z = Z 0 + 10 −6 [Ω X Y0 − Ω Y X 0 ](t − t 0 )

48
Table 3. 3 Cartesian rotation vector for each plate using the NNR-NUVEL1A kinematic
plate model (IERS Conventions (1996, p. 14). The units are radians per million years.)

Plate Name ΩX ΩY ΩZ
Pacific -0.001510 0.004840 -0.009970
Africa 0.000891 -0.003099 0.003922
Antarctica -0.000821 -0.001701 0.003706
Arabia 0.006685 -0.000521 0.006760
Australia 0.007839 0.005124 0.006282
Caribbean -0.000178 -0.003385 0.001581
Cocos -0.010425 -0.021605 0.010925
Eurasia -0.000981 -0.002395 0.003153
India 0.006670 0.000040 0.006790
Nazca -0.001532 -0.008577 0.009609
North America 0.000258 -0.003599 -0.000153
South America -0.001038 -0.001515 -0.000870
Juan de Fuca 0.005200 0.008610 -0.005820
Philippine 0.010090 -0.007160 -0.009670
Rivera -0.009390 -0.030960 -0.012050
Scotia -0.000410 -0.002660 -0.001270

3.5 Atmospheric Tides

The gravitational pull of the sun and the moon affects the solid Earth, the oceans, and the

atmosphere in different ways because of the properties of the material involved.

Atmosphere tides fundamentally affect the ocean and Earth tides in an indirect way.

Sea level is affected as a result of atmospheric pressure variations. Spatial and temporal

variations of atmospheric mass deform the Earth's surface. Many studies demonstrate possible

vertical displacements of up to 25 mm, but about one-third of this amount for horizontal

displacements. A simplified form in computing vertical displacement (mm) is (IERS

Conventions, 1996),

∆r = −0.35 p − 0.55 p 3. 5

49
where p is the local pressure anomaly with respect to the standard pressure (101.3 kPa) andp

is the averaged pressure anomaly within the 2000 km radius surrounding the site. Both

quantities have units of mbar (or 0.1 kPa).

50
4 International Terrestrial Reference Frame (ITRF)

Many reference systems and reference frames have been introduced and made available to the

public. Examples are the World Geodetic System (WGS-84), PZ-90, and the highly accurate

International Terrestrial Reference Frame (ITRF).

4.1 General Statements on Reference Frames

The Earth, as with all celestial bodies, is not static in nature. The Earth moves, rotates and

undergoes deformation. Since motion and position are not absolute concepts, they can be

mathematically described only with respect to some reference of coordinates (Kovalevsky and

Mueller, 1989) called a reference frame. According to Kovalevsky and Mueller (1989), the

purpose of a reference frame is to provide the means to materialize a reference system so that

it can be used for the quantitative description of positions and motions on the Earth (terrestrial

frames) or of celestial bodies including the Earth in space (celestial frames). In constructing

the reference frame, a set of parameters must be chosen. Thus the term ‘convention’ has been

used to characterize this choice. After defining the model in detail employed in the

relationship between its configuration of the basic structure and its coordinates, the

coordinates are thus thoroughly defined, but not necessarily accessible, hence the term a

conventional reference system. The term ‘system’ refers to the inclusion of the description of

the physical environment as well as the theories utilized in the coordinate definition. In

51
making a reference system available to users, it is normally materialized through a number of

points, objects, or coordinates, and a set of parameters. These define a conventional reference

frame. The reference frame must be accessible and clearly defined without ambiguity in

writing equations of motion of a body whose coordinates are referred to in the frame.

In 1988, the International Earth Rotation Service (IERS) was established at a site in

Paris under the cooperation of the International Astronomical Union (IAU) and the

International Union of Geodesy and Geophysics (IUGG). IERS assumed the responsibility of

the Bureau International de l’Heure (BIH). The goal of IERS is to provide to the worldwide

scientific and technical community the reference values for Earth orientation parameters and

reference realizations of internationally accepted celestial and terrestrial reference systems

(LAREG, 2000a).

The International Terrestrial Reference Frame (ITRF) was established and is

maintained by the Terrestrial Reference Frame Section of the Central Bureau (CB) of the

IERS. Currently, there are three products generated by the IERS CB including the ITRF, the

realization of the International Celestial Reference System (ICRS) (a space-fixed system that

refers to the positions of quasars and other celestial objects), and the determination of Earth

orientation parameters (EOP) (i.e., Universal Time, nutation corrections, and polar motion

coordinates) which relate the ITRS and the ICRS.

The ITRF implementation was originally based on the combination of Sets of Station

Coordinates (SSC) and velocities derived from observations of space-geodetic techniques such

as Very Long Baseline Interferometry (VLBI), Lunar Laser Ranging (LLR), and Satellite

Laser Ranging (SLR). IERS augmented the methodology to include GPS in 1991 and the

Doppler Orbitography and Radio-positioning Integrated by Satellite (DORIS) in 1994

(Boucher et. al., 1996). IERS regularly performs annual ITRF solutions, which are published

52
in the IERS Annual Reports and Technical Notes. Since 1988, IERS has evolved many ITRF

solutions, namely ITRF-97, 96, and 94 to 88. The on-going ITRF effort is called the ITRF-

2000 (LAREG, 2000d) which includes not only active space geodetic instruments, but also

useful markers. In addition, the IGEX-98 GLONASS stations are expected to be part of the

ITRF-2000 solution.

4.2 The ITRF

The ITRF represents the International Terrestrial Reference System (ITRS). ITRS has an

origin at the mass center of the whole Earth that takes the oceans and the atmosphere into

account (Boucher and Altamimi, 1996; LAREG, 2000a). The ITRS is realized by estimates of

the coordinates and velocities of a set of observing stations of the IERS. The ITRS uses

International Standard (SI) meter for its length unit defined in a local Earth frame in the

meaning of a relativistic theory of gravitation. According to the resolutions by the IAU and

the IUGG, the orientation of the ITRS axes is consistent with that of the BIH System at 1984.0

within ± 3 milli-arc-second (mas) and the time evolution in orientation of ITRS has no

residual rotation relative to Earth’s crust (Boucher and Altamimi, 1996).

4.3 Transformation Between ITRFs

Presently, in the ITRF computation process, the following basic procedure has been

implemented (Boucher and Altamimi, 1996):

? Reduction of each individual SSC’s at a common reference epoch t0 using their

respective station velocity models (fixed geophysical plate motion models or

estimated velocity fields),

53
? At a reference epoch t0, the least-squares estimation yields ITRF station

coordinates, in addition to seven transformation parameters, for each SSC with

respect to the ITRF. The combination procedure uses the standard model based

on Euclidean similarity involving the seven parameters with the general

transformation form between ITRF coordinates (X, Y, Z) and individual solution

coordinates (Xs, Ys, Zs) having three respective translations: T1, T2, and, T3; three

respective rotations: R1, R2, and R3; and the scale factor: D.

 Xs   X   T 1   D − R3 R 2   X 
 Ys  =  Y  + T 2 +  R3 D − R1  Y  4. 1
      
 Zs   Z  T 3 − R 2 R1 S   Z 

? Appropriate variance is assigned for local ties between co-located stations

? The ITRF velocities have been estimated either by

1. Combination, similar to the procedure used in the combination of station

coordinates; therefore, the method of combining velocities is equivalent to

and consistent with the method for combining station coordinates and can

make use of derivatives from the model, or

2. Differentiating combined coordinates at two different epochs.

Table 4.1 provides parameters from ITRF-94 to previous ITRF series, published in

previous IERS technical notes. From equation (4.1), (X, Y, Z) are the coordinates in ITRF-94

and (Xs, Ys, Zs) are the coordinates in the other frames. Rates must be applied for ITRF-93.

By construction, the transformation parameters between ITRF-94, ITRF-96 and ITRF-97 are

zero (Altamimi, 2000). The time epoch is used to indicate the position in a time series of

ITRF.

54
Table 4. 1 Transformation parameters from ITRF-94 to other ITRFs
(McCarthy, 1996; Altamimi, 2000)

4.4 Orientation and Origin of the ITRF

4.4.1 Orientation

From versions ITRF-88 through ITRF-92, the orientation was defined such that no rotation

existed between these frames. However, the orientation of ITRF-93 was constrained to be

consistent with the IERS series of Earth Orientation Parameters at epoch 88.0. The ITRF-94

orientation is again constrained to be consistent with the ITRF-92 at epoch 1988.0. For ITRF-

96 and ITRF-97, the reference frame definition (origin, scale, orientation, and time evolution)

of the combination is achieved in such a way that ITRF-96 and ITRF-97 are in the same

system as the ITRF-94 (LAREG, 2000b; LAREG, 2000c).

55
4.4.2 Origin

The ITRS origin is located at the center of mass of the whole Earth, including the oceans and

the atmosphere. The origins from series ITRF-88 through ITRF-92 were fixed to the

respective ICRS SLR solutions included in each ITRF calculation.

4.5 The Draft ITRF-2000 Reference Frame

ITRF-2000 is an adoption of the newest ITRS realization. ITRF-2000 consists of not only a

set of positions and velocities of global network tracking stations, but also related useful

markers recognized by a wide application community such as geodesy, cartography, etc.

ITRF-2000 includes the previous ITRF stations and expands to cover other types of points:

• PRARE stations,

• IGEX-98 GLONASS stations,

• points located at tide gauges, following the GLOSS (Global Sea Level Observing

System) and related programs such as EOSS (European Sea-level Observing

System) or EUVN (European Vertical Reference Network)

• points linking high accuracy gravity sensors or time/frequency laboratories

• calibration sites for satellite altimetry, and

• markers useful for national surveying agencies.

The quality of each individual point is specified according to the ITRF-2000 quality

criteria guidelines. In particular, the stations fulfilling the International Space Geodetic

Network (ISGN) criteria are identified in the publication. In addition, a validation process has

been established, which had not been applied in all previous ITRF realization processes.

56
4.6 GPS WGS-84

The military reference frame World Geodetic System 1984 (WGS-84) is applied to the GPS

system. The WGS-84 Coordinate System is a Conventional Terrestrial Reference System

(CTRS) utilizing a right-handed Earth-fixed orthogonal coordinate system. Its z-axis is in the

direction of the IERS Reference Pole (IRP) that corresponds to the direction of the BIH

Conventional Terrestrial Pole (CTP) at epoch 1984 with an uncertainty of 0.005” (IERS

Conventions, 1996, page 11; NIMA, 1997, page 2-2). Its x-axis is the intersection of the IERS

Reference Meridian (IRM) with the plane passing through the origin and normal to the z-axis

(NIMA, 1997, page 2-2). The IRM is coincident with the BIH Zero Meridian) at epoch 1984

with an uncertainty of 0.005” (IERS Conventions, 1996, page 11). Its y-axis completes a

right-handed ECEF orthogonal coordinate system. The latest realization of WGS-84 frame is

at epoch 1997 and the name has been given as WGS-84(G873). This realization is

implemented by the National Imagery and Mapping Agency (NIMA). The letter ‘G’ indicates

that the observation coordinates were obtained through GPS techniques and that Doppler data

were not included in the observations materializing the frame. The number ‘873’ is the GPS

week number at epoch 0h UTC on 29 September 1996 of a first date when coordinate frame

was made available through NIMA GPS ephemerides. The WGS-84(G873) represents

NIMA’s latest geodetic and geophysical modeling of the Earth from a geometric, geocentric,

and gravitational standpoint based on data, techniques, and technology available through 1996

(NIMA, 1997, page 1-1). This is the third edition of WGS-84. The previous versions of WGS

reference frames are WGS-84 and WGS-84(G730), which were implemented in the NIMA

GPS precise ephemeris estimation process ranging from 1 January 1987 to 1 January 94 and 2

January 94 to 28 September 1996, respectively. The station coordinates which compose the

operational WGS-84 reference frame are those of the permanent DoD GPS monitor stations.

WGS-84(G873) was implemented in the GPS Operational Control Segment (OCS) and

57
incorporated into the Kepler elements of the broadcast message on 29 January 1997. The

WGS-84 origin serves as the geometric center of the WGS-84 Ellipsoid, and the z-axis serves

as the rotational axis of this ellipsoid of revolution (NIMA 1997, page 3-1).

4.7 Agreement Between WGS-84 and ITRF

A comparison of coordinates between two reference frame systems can be made after the

adjustment of a best fitting seven-parameter transformation. Malys and Slater (1994) reported

an agreement between WGS-84(G730) and ITRF-92 at the 0.1 m level. Daily comparisons of

WGS-84(G873) and ITRF-94 through their respective precise orbits reveal systematic

differences no larger than 2 cm. The day-to-day dispersion on these parameters indicates that

these differences are statistically insignificant (NIMA, 1997, page 2-5).

58
5 Troposphere and Ionosphere

Satellite signals travel through the atmosphere which affects the state of the signals. These are

divided into two effects, tropospheric and ionospheric. Each effect influences the satellite

signals differently. Since the troposphere is a non-dispersive medium, tropospheric refraction

causes an identical effect on both code and phase modulation. The troposphere causes a signal

delay of up to 30 meters for a horizontal path. Therefore, the effect from the troposphere is

considered one of the major sources of errors imposed on the satellite signals. On the other

hand, the ionosphere is a dispersive medium of the ionized atmosphere layer(s). Thus, the

ionosphere affects the signal code and phase modulation in an opposing way. Moreover, the

ionospheric effect is a function of carrier frequency. Fortunately, the ionospheric effect can be

eliminated via dual frequency observation. The environments of the troposphere and

ionosphere are described in Chapter 1 (see section 2.4.2 ).

In this chapter, the emphasis will be given to the recent tropospheric model or so-called global

mapping function. The global mapping function must be utilized for the global positioning

analysis. For the ionospheric effect, the popular ionosphere-free linear combination

expression will be given.

59
5.1 Standard Atmosphere

In the 1920s, the first modern standard atmosphere definition was developed in the U.S. by the

National Advisory Committee for Aeronautics (NACA) and in Europe by the International

Commission for Aerial Navigation (ICAN) to fulfill a need to standardize aircraft instruments

and improve flight performance. Theoretical aspects of the physics of the air were utilized to

build the standard atmosphere. In 1952, the discrepancy between the two independently

developed NACA and ICAN standards was eliminated through the adoption of a new standard

atmosphere definition by the International Civil Aviation Organization (ICAO) with computed

tables ranging from 5 km below to 20 km above mean sea level. The U.S. Committee on

Extension to Standard Atmosphere (COESA) was formed in 1953, and in 1961 a working

group was convened to define a new standard atmosphere up to an altitude of 700 km. The

work of COESA led to the new versions of the U.S. Standard Atmosphere with slight

modifications from those previously adopted. These models were published jointly by the

National Oceanic and Atmospheric Administration (NOAA), the U.S. Air Force, and NASA.

The U.S. Standard Atmosphere 1976 has been used until now. The standard atmosphere is

essentially defined in terms of an ideal air obeying the perfect gas law and by assumption that

the atmosphere is static with respect to the earth (Laurila, 1976). It is based on the standard

values for air density, temperature, and pressure at sea level as given in the Table 5.1.

Table 5. 1 Sea level standard values

Temperature, TSea 288.15 K

Pressure, Psea 101325 Pa (or N/m2) (1013.25 mb)

Gas constant, R 8.31432 x 103 Nm/(kmol K)

Density, ρsea 1.225 kg/m3

Gravity acceleration, gsea 9.80665 m/s2

60
In addition to perfect gas theory, rocket and satellite data atmospheric pressure, density, and

temperature were used to represent the Earth's atmosphere from sea level to 1000 km. Single

profiles representing the idealized, steady-state atmosphere for moderate solar activity are

applied for the U.S. Standard Atmospheres 1958, 1962, and 1976 (NOAA/NASA/USAF,

1976; NASA, 2000a). Below 32 km the U.S. Standard Atmosphere agrees with the ICAO

standard atmosphere for all practical purposes. However, the U.S. Standard Atmosphere does

not necessarily represent an average of the vast amount of atmospheric data today from

observations within that height region, particularly for heights below 20 km. Parameters listed

include temperature, pressure, density, acceleration caused by gravity, pressure scale height,

number density, mean particle speed, mean collision frequency, mean free path, mean

molecular weight, sound speed, dynamic viscosity, kinematic viscosity, thermal conductivity,

and geopotential altitude. The altitude resolution varies from 0.05 km at low altitudes to 5 km

at high altitudes. Units in all tables are given in English (foot) as well as metric (meter) units.

The U.S. Standard Atmosphere Supplements (1966) includes tables of temperature, pressure,

density, sound speed, viscosity, and thermal conductivity for five northern latitudes (15, 30,

45, 60, and 75 degrees), for summer and winter conditions (NASA, 2000a), which departs

from the U.S. Standard Atmosphere.

The U.S. Standard Atmosphere utilized the linearly segmented temperature high

profile, and the assumption of hydrostatic equilibrium, in which the air is treated as a

homogenous mixture of the several constituent gases. The fundamental seven layers of the

Standard Atmosphere (1976) from sea level to 86 km (Table 5.2) includes (geopotential)

height and temperature gradient by altitude (temperature lapse rate).

61
Table 5. 2 The fundamental seven layers of the U.S. Standard Atmosphere 1976

H1 (km) H2 (km) α = dT / dh
Layer
From To (°K/km)

1 0 11 −6.5
2 11 20 0.0
3 20 32 1.0
4 32 47 2.8
5 47 51 0.0
6 51 71 −2.8
7 71 84.852 −2.0

Based on the standard sea level values given in Table 5.1, the atmospheric parameters can then

be computed using information given in Table 5.2. The temperature variation can be

expressed as a sum of a series of lower layers, linear altitude. The temperature T at height h

(in km) falling in layer n is written as:

n −1
T = TSea + α n (h − H1, n ) + ∑ α i ( H 2, i − H 1, i ) 5. 1
i =1

where TSea is the temperature at sea level. In the range of each atmosphere layer where

temperature varies linearly as a function of altitude, the pressure can be calculated from the

following expression:

g M
− 0
 α n (h − H 1,n ) + T  αn R
P = Pn ⋅   , for α ≠ 0 5. 2a
 T 

g M 
P = Pn exp 0 (h − H 1,n )  , for α = 0 5. 2b
 RT 

62
where Pn is the pressure at the initiative point of each altitude range, gravity

g 0 = 9.80665 m/s2, and universal gas constant R is given Table 5.1. Having pressure and

temperature, the density at a specific altitude can be calculated from

PM
ρ= 5. 3
RT

in which M is a constant mean molecular weight of the gas.

At greater heights (i.e. between 86 and 1000 km), the definitions governing the

Standard are far more complex due to dissociation and diffusion processes producing

significant departure from homogeneity. Therefore, the temperature height profile cannot be

expressed as a series of linear functions like those employed at lower attitudes. Relevant

expressions and standard tables are given in NOAA/NASA/USAF (1976).

The variations of gravity acceleration g defined by U.S. Standard Atmosphere through

manipulating the inverse-square law of gravitation, with sufficient accuracy for most model

atmosphere computations is

2
 r 
g = g 0  0  5. 4
 r0 + Z 

where r0 = 6,356,766 m is the effective radius of the earth at sea-level in which the centrifugal

acceleration is taken into account, Z is the geometric height at a specific latitude, and g0 =

9.80665 m/s2. For the U.S. Standard Atmosphere, this equation is valid from sea level to the

geometric altitude of 1000 km.

63
5.2 Troposphere

The neutral atmosphere, which is the non-ionized part of atmosphere, can normally be divided

into two components, the hydrostatic (dry) and wet portions of the troposphere. The

hydrostatic component consists of mostly dry gases (normally referred to the dry part) ,

whereas the wet component is a result of water vapor. The troposphere causes radio signal

delay. The hydrostatic fraction contributes approximately 90% of the total tropospheric

refraction (Leick, 1995, page 308). For high accuracy positioning, correcting the delay of

radio signals as they traverse the neutral atmosphere is necessary as it is one of the dominant

error sources. Tropospheric effect is frequency-independent and cannot be eliminated via

dual-frequency observations. The tropospheric path delay can be defined as (e.g., Janes et al.,

1991; Mendes, 1994):

 
Tkp = ∫ [ n ( r ) − 1] csc θ ( r ) dr + 
 ∫ csc θ ( r ) dr − ∫ csc ε ( r ) dr 

5. 5
path  path path 

where r is geocentric radius, n is the refractive index, and θ and ε , respectively, refer to

refracted (apparent) and non-refracted (geometric or true) satellite elevation angle; n relates to

the tropospheric refractivity N Trop as given below:

N Trop = (n − 1) ⋅10 6 5. 6

Note that equation (5.5) holds for a spherically symmetric atmosphere, and n is

allowed to vary along the signal path as a function of geocentric radius. The first term

characterizes the deviation of electromagnetic path s from geometric length of the refracted

transmission path. The bracketed term is the geometric delay accounting for path curvature

(ray bending), which is the difference in the geometric lengths of the electromagnetic and

rectilinear paths from the satellite to the observing station (Janes et al., 1991). Such curvature

64
effect is essentially significant for satellite elevation angles of 10-20 degrees; therefore, in

practice, the bracketed term is often omitted. For satellite signals, we obtain tropospheric

delay:

Tkp = 10 −6 ∫N
Trop
ds 5. 7
path

Allowing for the hydrostatic and the wet components, tropospheric delay can be rewritten as:

Tkp = 10 −6 ∫ (N + N wTrop ) ds
Trop
d
path
5. 8
∫ N d ds + 10 ∫ N w ds
−6 −6
= 10 Trop Trop

path path

where N dTrop and N wTrop correspond to tropospheric refractivity of the hydrostatic and the wet

components. In 1974, Thayer expressed the refractivity NT in term of absolute temperature

and partial pressure of the dry gases (Pd) and of water vapor ( e0 ) in millibars, viz. (e.g.,

Mendes et al., 1994):

Pd −1  e0 e 
N T = K1 Z d + K 2 + K 3 02  Z w−1 5. 9
T0  T0 T0 

where the constant coefficients K1, K2, and K3 are empirically determined. T0 is absolute

temperature in Kelvins at the tracking station. Zd and Zw are corresponding compressibility

factors for dry air and water vapor, which account for the departure of the air behavior from

that of the ideal gas and rest on the partial pressure due to dry gases and temperature. The first

term on the right side of equation (5.9) refers to N dTrop , whereas the bracketed terms refer to

N wTrop . The frequently used sets of refractivity constants are given in Table 5.3.

65
Table 5. 3 Frequently used refractivity constants
(e.g., Bean et al., 1966; Mendes et al., 1994; Langley, 1996)

Refractivity Smith and Weintraub Thayer


coefficients [1953] [1974]
K1 (K/mb) 77.61±0.01 77.604 ± 0.014
K2 (K/mb) 72 ± 9 64.79 ± 0.08
K3 (K2/mb) (3.75 ± 0.03) 105 (3.776 ± 0.004) 105

Normally the vertical wet and dry refractions are related to the refraction of a particular

elevation angle by the mapping function. The wet component depends on water vapor

content. Much research has been focused on modeling water vapor content.

The water vapor pressure e0 can be calculated from a priori knowledge of

environmental information such as relative humidity and temperature at the tracking station.

Water vapor pressure in millibars recommended in the IERS Conventions (1996) is

7.5(T0 − 273.15)
237.3+T0 − 273.15
e0 = 0.0611 RH 10 5. 10

where RH is the relative humidity at the observing station in percent.

5.2.1 Tropospheric Models

Due to the significance of tropospheric effects on radio signal propagation, many studies to

formulate tropospheric correction have been performed. Various tropospheric models exist.

In addition, different mapping functions which illustrate signal delay as a function of elevation

angle are also given. The following provides descriptions of frequently used tropospheric

models and mapping functions.

66
5.2.1.1 Hopfield Model

Hopfield empirically developed a tropospheric model in 1969 using worldwide data. The

Hopfield model applies a single layer polytropic model atmosphere ranging from the Earth's

surface to altitudes of about 11 km and 40 km for the wet and dry layers, respectively

(Hopfield, 1969; Janes et al. 1991; Hofmann-Wellenhof et al., 1997), see Figure 5.1.

Figure 5. 1 Hopfield single-layer polytropic model atmosphere

The Hopfield model shows dry and wet refractivity components as a function of tracking

station height h above the Earth's surface and is given in the following forms:

µ
Hd
− h
N dTrop = N dTrop
,0   5. 11a
 Hd 

µ
Hw
− h
N wTrop = N wTrop
,0   5. 11b
 Hw 

where

67
µ=4 empirically determined power of the height ratio,

Hd = 40136 + 148.72(T − 273.16) a polytropic thickness for the dry part (m),

Hw = 11000 a polytropic thickness for the wet part (m),

P0
,0 = K1
N dTrop dry tropospheric refractivity for the station at the
T0

Earth's surface as a function of pressure (millibars)

and temperature (Kelvin),

e0 e
,0 = K 2
N dTrop + K 3 02 wet tropospheric refractivity for the station at the
T0 T0

Earth's surface as a function of water vapor, pressure,

and temperature.

Inserting equation (5.11) into equation (5.8), and integrating each element with the respective

integration ranges along the vertical direction (i.e. from h = 0 to h = Hd and from h = 0 to h =

Hw for the dry and wet components), we then obtain tropospheric zenith delay in units of

meters:

T =
k
Z

5
[
10 −6 Trop
N d , 0 H d + N wTrop ]
,0 H w . 5. 12

The values of dry and wet polytropic thickness, Hd and Hw, are typically in the range of 40-45

km and 10-13 km, respectively (Hofmann-Wellenhof et al., 1997). Hopfield's zenith

tropospheric delay equation (5.12) can be employed together with a mapping function to

obtain tropospheric delay at a specific satellite elevation angle.

68
5.2.1.2 Saastamoinen Model

Saastamoinen (1971) applied the law of Gladstone and Dale; that the height integral

∫ ( n − 1) dr of the atmospheric refractivity for radio microwaves taken from ground level to

the top of the stratosphere is, in a dry atmosphere, directly proportional to the ground pressure.

Derivation of the Saastamoinen model involves thinking of the atmosphere as a mixture of

two ideal gases, dry air and water vapor. Gas laws are then applied to derive refractivity. The

temperature in the troposphere from sea level to about ten kilometers decreases with height at

a fairly uniform rate which varies slightly with latitude and season, although in the polar

regions there is a permanent inversion in the lower troposphere where the actual temperatures

initially increase with height.

Saastamoinen divided the dry atmosphere into two layers: a polytropic troposphere

extending from the surface to an altitude of approximately 11-12 km, and an isothermal

stratosphere continuing from the troposphere to approximately 50 km as shown in Figure 5.2 .

Atmospheric water vapor is confined to the troposphere only.

For the normal mid-latitude conditions, the Saastamoinen model is given in units of

meters as:

  1255  
Tkp = 0.002277 sec z  P0 +  + 0.05 e0 − Ω tan 2 z  5. 13
  T0  

where z is the true zenith distance, P0 is the pressure at the observed station in millibars, and

the coefficient

R  P0 T0 − ( Rβ / g ) p 0T 0 
Ω=  . 5. 14
rg 1 − Rβ / g 

69
Figure 5. 2 Schematic of Saastamoinen tropospheric and stratospheric spherical layered

dry atmosphere

R is the gas constant, r is the earth’s radius, T0 is temperature at the tracking station, and p0

and T0 are pressure and temperature at the bottom of the stratosphere. β is the vertical

gradient of temperature ( dT / dh ). Figure 5.3 shows a plot of the coefficient Ω varied with

station height above sea level as given by Saastamoinen (1971).

For a station at sea level, Ω is approximately 1.16. For the signal coming in the zenith

direction (z = 0), the term involving coefficient Ω is zero, and equation 5.13 can be rewritten

for tropospheric zenith delay as:

  1255  
TkZ = 0.002277  P0 +  + 0.05 e0  5. 15
  T0  

70
Coefficient for Saastamoinen Model
1.2
1.1

1
0.9
coefficent


0.8

0.7
0.6

0.5
0.4
0 1 2 3 4 5 6
h
Station height km above sea level

Figure 5. 3 Coefficient Ω for Saastamoinen model versus height

The first term is a function of surface pressure and refers to the hydrostatic component, while

the rest is a function of water vapor pressure and corresponds to the wet component. A

correction to standard gravity at the observing station has been incorporated into the

Saastamoinen model (Janes and Langley, 1989; Janes et al., 1991):

0.002277   1255  
TkZ =  P0 +  + 0.05 e0  5. 16
g′   T  

where

g ′ = 1 − 0.0026 cos 2ϕ − 0.00028 h0 5. 17

has units m/s2, and ϕ and h0 are station latitude and orthometric height (km) (Saasamoinen,

1971). Saastamoinen estimates the accuracy of the hydrostatic and the wet components as 2-3

mm and 3-5 mm rms, respectively. Mendes (1998) found that the hydrostatic component can

be predicted with sub-millimeter accuracy from the Saasamoinen model if accurate

measurements of surface pressure are available. For the wet component he found that it is

71
much more difficult to use the surface meteorological data to predict with the best models

show rms of a few centimeters .

5.2.2 Mapping Functions

The mapping function describes the elevation angle dependence of the delay of the signals that

travel through the neutral atmosphere (Niell, 1996). Each neutral atmospheric component, the

zenith delay and a mapping function are used to model the line of sight delay. Azimuthal

symmetry is usually assumed. The mapping function parameters are normally involved with

temperature, pressure, and relative humidity. Moreover, some mapping function models, e.g.

MTT by Herring (1992) and NMF by Niell (1996), take latitude and height above sea level

into account. An extensive comparison between mapping functions and ray traces of

radiosonde profiles is given in Mendes (1998).

Generally, the tropospheric delay correction is defined in terms of the contributions of

the hydrostatic and wet components. Therefore, the total tropospheric delay can be described

by:

τ a ( E ) = τ hz mh ( E ) + τ wz m w ( E ) 5. 18

where the total tropospheric delay τa for unrefracted observation elevation angle E is being

considered. The symbols τ hz ,τ wz , mh , and m w are the zenith delays, and the mapping

functions for the hydrostatic and wet atmospheric components, respectively. The nominal

zenith delays for a site at sea level are 2300 mm for the hydrostatic and 100 mm for the wet

component. Wet zenith delay extreme values are 300 and 400 mm for midlatitude and tropical

regions, respectively (Niell, 1996). However, the more specific values of zenith delay may be

obtained from the tropospheric models.

72
Because of the atmospheric curvature, the mapping function does not just change as

the cosecant of the elevation angle E, which would be expected for a plane parallel refractive

medium. The hydrostatic mapping function will change according to changes in the ratio of

the atmospheric thickness to the Earth’s radius, as a result of the changes in temperature

(Niell, 1996). Therefore, the temporal change of mapping function is associated with

variability of temperature at various atmospheric heights, which can be obtained from

radiosonde profiles.

5.2.2.1 Marini Mapping Function

In 1972, J. W. Marini developed a tropospheric correction, which shows that the elevation

angle E dependence of any horizontally stratified atmosphere can be approximated by

expanding in a continued fraction in term of 1 / sin E . The general form of the Marini

mapping function is written as (Marini, 1972):

1
m( E ) = 5. 19
a
sin E +
b
sin E +
c
sin E +
sin E + K

E is the geometric (unrefracted) satellite elevation angle and a, b, c,... are profile dependent

coefficients. The Marini mapping function does not explicitly separate the hydrostatic and

wet components of tropospheric delays.

5.2.2.2 Marini & Murray Mapping Function

The Marini & Murray mapping function maps the total delay based on Saastamoinen zenith

delay viz.(Marini and Murray, 1973):

73
1+ ζ
m( E ) = 5. 20
ζ
1+ζ
sin E +
sin E + 0.015

wherein

G
ζ = , 5. 21
TkZ

0.002644 [−0.14372 h0 ]
G= e 5. 22
g′

Surface meteorology information as well as station altitude and latitude are incorporated into

the Marini & Murray mapping function. TkZ and g ' are previously given in equations (5.16)

and (5.17). The Marini & Murray formula is specified to be valid for an elevation angle

greater than 10°. In addition, the comparison of the Marini & Murray formula with ray traces

of radiosonde data revealed that the standard deviation of the range correction increases from

20 mm at the zenith to nearly 200 mm at 10°.

5.2.2.3 Chao Mapping Function

In 1974, C. C. Chao derived a tropospheric mapping function to be used for radio tracking

corrections of the 1971 Mariner Mars spacecraft. Chao treats the wet and dry components

separately through empirical fitting to an average refractivity profile derived from two years of

radiosonde data (Chao, 1974).

74
1
mh ( E ) = 5. 23a
ah
sin E +
tan E + bh

1
mw ( E ) = 5. 23b
aw
sin E +
tan E + bw

where a h = 0.00143, bh = 0.0445, a w = 0.00035, and bw = 0.0170 . It should be

noted that the term tan(E ) was used to ensure that both dry and wet mapping functions are

one at the zenith. The accuracy of the Choa's dry mapping function is 1% down to 1° with

respect to the ray trace of the average annual refractivity profiles. The error at this level is too

large for accurate geodetic VLBI application (Niell, 1996). Due to vast spatial and temporal

variability, Choa's wet mapping function was sufficiently accurate for space geodetic

measurements until the introduction of the Herring mapping function (Niell, 1996).

5.2.2.4 Lanyi Mapping Function (Lanyi)

Lanyi (1984) constructed a total mapping function that has both the dry and wet components.

The Lanyi mapping function was expected to be applicable to a 6° minimum elevation angle.

The parameterization of the Lanyi mapping function includes surface temperature, height of

isothermal surface layer, and temperature lapse rate. The greater parameterization in

temperature profiling of the Lanyi mapping function allows inclusion of an isothermal layer

of variable height beginning at the surface. The IERS Conventions (1996) prefers the Lanyi

mapping function if information about the vertical temperature distribution in the atmosphere

is available.

75
5.2.2.5 Davis Mapping Function (CfA-2.2)

Both the Marini & Murray (1973) and Chao (1974) mapping functions are in the generalized

forms, and thus are exposed to the influence of variability in the refractivity profile and lateral

gradients. To increase accuracy at low elevation angles and to better accommodate local and

seasonal variations, further modification was performed by Davis et al. (1985) by adding the

fraction of the sine term to Chao's mapping function. The Davis mapping function was

dubbed "CfA-2.2":

1
mh (ε ) = 5. 24
a
sin ε +
b
tan ε +
sin ε + c

in which through the least-square fit the ray trace yields coefficients as linear functions of the

surface weather conditions, i.e. pressure, temperature, and relative humidity.

a = 0.001185 [ 1 + 6.071 ⋅ 10 −5 ( P0 − 1000) − 1.471 ⋅ 10 −4 e0 + 3.072 ⋅ 10 −3 (T0 − 20)


+ 0.01965(α + 6.5) − 0.005645( H t − 11.231)]

5. 25

b = 0.001144 [ 1 + 1.164 ⋅ 10 −5 ( P0 − 1000) − 2.795 ⋅ 10 −4 e0 + 3.109 ⋅ 10 −3 (T0 − 20)


+ 0.03038(α + 6.5) − 0.001217( H t − 11.231)]

5. 26

c = −0.0090 5. 27

From above, H t is the height of the tropopause (km). α is the tropospheric temperature lapse

rate value. Even though Davis et al. (1985) evaluated parameters a and b by least-squares fit

76
to ray traces of idealized pressure, temperature, and humidity profiles of spherical symmetric

layered atmosphere, they remarked that errors of 1 to 2 mm are present for elevation angles

from 20° to 60°. This is due to the incorrectness of the tangent term since tan E does not

approach sin E quickly enough. However, the advantage of CfA-2.2 is its simplicity, both in

calculating the mapping function itself and in calculating partial derivatives of the mapping

function with respect to the parameters to be estimated (Davis et al., 1985).

The tropospheric temperature lapse rate (α) normally has values ranging from −6 to

−7°K/km, but −6.5°K/km is the standard value according the U.S. Standard Atmosphere

(NOAA/NASA/USAF, 1976). However, Mendes and Langley (1998) reported a mean global

value of 6.17 ± 0.82° / km . Mendes and Langley (1998) also observed a correlation between

α and surface temperature as expressed in the linear model as a function of temperature:

α (° / km ) = −5.930 − 0.0359(T0 − 273.16) 5. 28

The rms agreement of this mapping function (equation 5.24) compared with ray tracing is less

than 5 mm for all elevations above 5°. Mendes and Langley (1998) averaged the global

tropopause height data and reported H t = 11.3 ± 2.6 km, while Davis et al. (1985) suggested

a value of 11.231 km. The large standard deviation of Mendes and Langley (1998) reflects the

large latitudinal and seasonal variations, and therefore cannot be accounted for through a

single nominal number. The tropopause is highest at the equatorial regions with very small

seasonal variation and attenuates as it approaches polar areas where large seasonal variations

are detected from middle to high latitudes. Mendes and Langley (1998) found that tropopause

height is highly correlated with surface temperature according to the expression, in km

T0 − 273.16
Ht = 7.508 + 2.421 e 22.90 5. 29

77
5.2.2.6 Herring Mapping Function (MTT)

In 1992, T. A. Herring applied the Marini mapping function, but this correction has been

normalized to unity at zenith as given below (Herring, 1992):

a
1+
b
1+
m( E ) = 1+ c 5. 30
a
sin E +
b
sin E +
sin E + c

The coefficients of the physical quantities a, b, and c can be estimated from the least-

squares fits of m(E) to ray traces of idealized temperature and humidity profiles for different

values of pressure, temperature, etc. These coefficients depend linearly on surface

temperature, the cosine of the station latitude and the height of station above the geoid

(ranging from 0-1600 m).

5.2.2.7 Niell Mapping Function (NMF)

Recently, Niell (1996) proposed the new mapping function (NMF) based on temporal changes

and geographic location rather than on surface meteorological parameters. He argued that all

previously available mapping functions have been limited in their accuracy by the dependence

on surface temperature, which causes three dilemmas. All of these are because there is more

variability in temperature in the atmospheric boundary layer, from the Earth's surface up to

2000 m. First, diurnal alterations in surface temperature cause much smaller variations than

those calculated from the mapping functions. Second, seasonal changes in surface

temperature are normally larger than upper atmosphere changes (but the computed mapping

78
function yields artificially large seasonal variations). Third, the computed mapping function

for cold summer days may not significantly differ from warm winter days. For example,

actual mapping functions are quite different than computed values because of the difference in

lapse rates and heights of the troposphere.

The new mapping functions have been derived from temperature and relative

humidity profiles, which are in some sense averages over broadly varying geographical

regions. Niell (1996) compared NMF and ray traces calculated from radiosonde data spanning

about one year or more covering a wide range of latitude and various heights above sea level.

Such comparison was to ascertain the validity and applicability of the mapping function NMF.

Through the least-square fit of four different latitude data sets, Niell (1996) showed

that the temporal variation of the hydrostatic mapping function is sinusoidal within the scatter

of the data.

Table 5. 4 Coefficients of the hydrostatic NMF mapping function (Niell, 1996)

Latitude ϕi
Coefficients
15° 30° 45° 60° 75°
Average
a 1.2769934e-3 1.2683230e-3 1.2465397e-3 1.2196049e-3 1.2045996e-3
b 2.9153695e-3 2.9152299e-3 2.9288445e-3 2.9022565e-3 2.9024912e-3
c 62.610505e-3 62.837393e-3 63.721774e-3 63.824265e-3 64.258455e-3
Amplitude
a 0.0 1.2709626e-5 2.6523662e-5 3.4000452e-5 4.1202191e-5
b 0.0 2.1414979e-5 3.0160779e-5 7.2562722e-5 11.723375e-5
c 0.0 9.0128400e-5 4.3497037e-5 84.795348e-5 170.37206e-5
Height Correction
aht 2.53e-5
bht 5.49e-3
cht 1.14e-3

79
For the hydrostatic NMF mapping function, the parameter a at tabular latitude ϕ i at

time t from January 0.0 (in UT days) is given as:

 t − T0 
a(ϕ i , t ) = aavg (ϕ i ) + aamp (ϕ i ) cos 2π  5. 31
 365.25 

where T0 is the adopted phase, DOY28. The linear interpolation between the nearest a(ϕ i , t )

is used to obtain the value of a(ϕ , t ) . For parameters b and c, a similar procedure was

followed.

Table 5. 5 Coefficients of the wet NMF mapping function (Niell, 1996)

Latitude ϕ
Coefficients
15° 30° 45° 60° 75°
aw 5.8021897e-4 5.6794847e-4 5.8118019e-4 5.9727542e-4 6.1641693e-4
bw 1.4275268e-3 1.5138625e-3 1.4572752e-3 1.5007428e-3 1.7599082e-3
cw 4.3472961e-2 4.6729510e-2 4.3908931e-2 4.4626982e-2 5.4736038e-2

The coefficients for the wet NMF mapping function are shown in Table 5.5. No

temporal dependence is included in the wet NMF mapping function. Therefore, only an

interpolation in latitude for each parameter is required. Height correction associated with the

NMF is given as:

dm( E )
∆m( E ) = H 5. 32
dh

dm( E ) 1
= − f ( E , a ht , bht , c ht ) 5. 33
dh sin( E )

80
where f ( E , a ht , bht , c ht ) is a three-term continued fraction (equation 5.30). The parameters

aht, bht, cht as given in Table 5.4 were determined by a least-squares fit to the height correction

at nine elevation angles, and H is the station height above sea level.

Mendes (1998) analyzed the large number of mapping functions by comparing against

radiosonde profiles from 50 stations distributed worldwide (32,467 benchmark values). The

models that meet the high standards of modern space geodetic data analysis are Ifadis, Lanyi,

MTT, and NMF. He found that for elevation angle above 15 degrees, the models Lanyi, MTT,

and NMF yield identical mean biases and the best total error performance. At lower elevation

angles, Ifadis and NMF are superior.

5.3 Ionosphere

Ionospheric effect is a result of electromagnetic waves of GPS signals travelling through a

dispersive atmosphere to the antenna. The effect inversely varies with the square of frequency

f of the signals. Having dual-frequency observation, ionospheric range errors can be removed

from observation data. Major ionospheric effects correspond to rise and fall of the number of

sunspots − the solar cycle.

5.3.1 Spatial and Temporal Variations

Normally, at mid-latitudes the ionospheric effect on GPS signals can be negligible. On the

other hand, the ionospheric scintillation activity is becoming more significant at lower

latitudes, especially in the hours immediately after sunset (Knight and Finn, 1996). In

addition, the ionospheric effects rise and fall according to number of sunspots. The sunspot

cycle, which has a vast effect, was discovered in 1843 by Samuel Heinrich Schwabe. Around

81
1848, Johann Rudolph Wolf, Swiss astronomer and astronomical historian, confirmed

Schwabe's discovery of a cycle in sunspot activity through the use of previous records that

defined the cycle's length more accurately, to be an average of 11.1 years. Figure 5.4 shows

annual sunspot numbers since 1700.

Figure 5. 4 Sunspot count 1700-1800 (top), 1800-1900 (middle), and 1900-2000 (bottom)

(NOAA, 2000)

82
Wolf discovered a daily technique to measure solar activity by simply counting the

number of individual spots and groups of spots on the sun's surface. He introduced the term

"the Zurich relative sunspot number" (or Wolf's sunspot number), a value equal to the sum of

the spots plus 10 times the number of groups. This method is still used today and daily

observations of sunspots are averaged to find annual values. Wolf sunspot counts rise and fall

roughly every 11 years, and its cycle is asymmetrical with an average 4.3 years to rise from a

minimum to the maximum and another 6.6 years to drop to a minimum once again.

Figure 5. 5 Sunspot number prediction for cycle 23 (NASA, 2000b)

From the record, the largest annual mean number (190.2) occurred during 1957-58.

The peak in the current sunspot cycle (number 23) is approaching around the middle of 2000

(see Figure 5.5). Sunspot numbers can be obtained from the archive at NOAA (2000). Figure

83
5.6 depicts monthly variation of sunspot counts. Wolf also discovered that the sunspot cycle

coincided with disturbances in the Earth's magnetic field.

140

120

100
Counts

80

60

40

Monthly Sunspot Numbers

20
M 9

M 0
8

0
98

99

00
M 8

9
98

99
8

9
-9

-0
-9

-9

-0
-9

l-9

l-9
v-9

v-9
n-

p-

n-

p-

n-
ar

ar
ay

ay

ay
ar

Ju

Ju
No

No
Se

Se
Ja

Ja

Ja
M

Figure 5. 6 Monthly mean sunspot numbers (NOAA, 2000)

Some of the phase variations are transformed to amplitude and phase via diffraction

resulting in an irregular but rapid variation in amplitude and phase, called scintillation (Leick,

1995, page 297). In other words, ionospheric scintillation is caused by small-scale

irregularities in the ionospheric electron density, and can disturb the amplitude and phase of

traversing radio signals. Such effects can cause severe fades in the signal or rapid phase

gradients that exceed a receiver ability to hold “lock” on the signal; several cases of loss of

lock under such conditions have been reported (Nordwall, 1996).

84
Among others, solar ultraviolet (UV) activity is a major cause of ionospheric

turbulence. For GPS, the major ionospheric effects pertinent to solar UV activity are

ionospheric range delays and amplitude fading and phase scintillation effects. Direct

measurements of solar UV radiation cannot be made from the Earth's surface due to

atmospheric absorption, instead data from solar cycles collected over a 300 year period are

used. Another surrogate measure of this UV radiation is the solar radio flux at wavelength

10.70 cm; however, this method is less subjective than measurements of sunspot numbers

(Klobuchar and Doherty, 1998).

5.3.2 Ionospheric Range Delay

Ionospheric range delays are directly proportional to the total electron content (TEC), which

varies along the transmission path and can be defined as

TEC = ∫N
path
e ( s ) ds 5. 34

where Ne is the local electron density (electrons/m3). The TEC represents the total number of

free electrons contained in a column with cross-sectional area of 1-square meter along the path

of signal between satellite and receiver. The TEC is in units of (el/m2). The Total Electron

Content Unit (TECU) is defined as TECU = 1 ⋅ 1016 el / m 2 . Transforming the time delay of

a code sequence or the phase advancement to the corresponding distance (in meters) is:

40.28 40.28
I kp, f , P =
f 2
TEC =
f 2 ∫ N e ds
path
5. 35
40.28 c
=
f 2 ∫ N e dt
path

85
The above equation is the ionospheric range delay or advance between receiver k and satellite

p for the carrier frequency f and c is the speed of light. The corresponding time delay or

advance follows as

I kp, f , P 40.3 TEC


vf = = 5. 36
c cf 2

Figure 5.7 shows GPS ionospheric range errors as functions of TECU and frequency.

GPS Ionospheric Range Error

30 100
I(f1) I(f2) 90
25 80
Ionosphere (m)

Time delay (ns)


20 70
60
15 50
40
10 30
5 20
10
0 0
0 25 50 75 100
TECU

Figure 5. 7 GPS Ionospheric range errors as functions of TECU and frequency

For GPS signals, it is necessary to identify the delays of the P1 and P2-codes and the

advances of the L1 and L2 carrier phases. Normally, the ionospheric code delay has a unit in

meters while carrier phases have a unit in cycles, except when the carrier phases have been

scaled to distance.

c p
I kp,1, P = − I kp,1, Φ = − I
f1 k ,1,ϕ
5. 37
c
I kp, 2, P = − I kp, 2, Φ = − I kp, 2,ϕ
f2

86
Ionospheric relations between two frequencies can be formed, for code and phase:

I kp,1, P f 22
= 5. 38
I kp, 2, P f12

I kp,1,ϕ f2
= 5. 39
I kp, 2,ϕ f1

5.3.3 Ionosphere Models

There are a few ionospheric models available to estimate ionospheric effect. Examples are the

ionospheric plate model, daily cosine model, and ionospheric point model. Ionospheric

coefficient for the cosine model included in the navigation message compensates

approximately 50% of the actual group delay. Details of these models are summarized in

Leick (1995, pages 299-302).

5.3.4 Functions of Observables

The effects from the ionosphere on GPS analysis can vary depending on many factors. Such

factors include geomagnetic variations, spatial locations, upper atmospheric chemical

composition and temperature, wind circulation, duration of the sunspot cycle, season, time of

the day, and line of sight. Since ionospheric effect is a function of a signal's frequency, having

dual frequency data can eliminate almost all of the ionospheric effects. In addition, an

extremely precise measurement of relative TEC can be formed from the linear combination of

the two carrier phases of two signals (Musman et al., 1998). Even the most dominant

ionospheric correction is the lowest order term ( 1 / f 2 ), the higher order terms 1 / f 3 and

87
4
1/ f might be important by affecting the observation accuracy at a few centimeters and

millimeters respectively (Bassiri and Hajj, 1992).

5.3.4.1 Dual-Frequency Ionosphere-free

Exploiting the ionospheric frequency relation equations (5.38) and (5.39), a combination to

obtain ionosphere-free functions for codes and phases can be formed (Leick, 1995; Langley,

1996):

f12 f 22
Pkp, IF (t ) = Pkp,1 (t ) − Pkp, 2 (t ) 5. 40
f12 − f 22 f12 − f 22

f12 f1 f 2
ϕ kp, IF (t ) = ϕ kp,1 (t ) − ϕ kp, 2 (t ) 5. 41
f12 − f 22 f12 − f 22

In equation (5.41), the ambiguity term is not an integer due to scaling factors employed to

eliminate the ionospheric term.

5.3.4.2 Dual-Frequency Ionosphere

Pseudorange linear combinations for the ionospheric solution can be established directly:

PI p, k = P1,pk − P2p, k

= (1 − α ) I Pp, k + c (1 − α ) TGD
p
5. 42

+ d1, P , k + d1p, P, k + d1p, P − d 2, P , k − d 2p, P , k − d 2p, P

where α = ( f1 / f 2 ) 2 . The receiver clock cancels in this combination. The TGD term is a

constant over a period of time, as given in the broadcast navigation message. Thus, this

88
ionospheric combination contains the difference of hardware delays and multipath that

normally are small magnitudes compared to the ionospheric effect, see Figure 8.8.

The dual-frequency phase ionospheric function can be written as:

f1 p
ϕ kp, I (t ) = ϕ kp,1 (t ) − ϕ (t )
f 2 k ,2
5. 43
N k , 2 (t ) − 1 (1 − α )I kp,1, P (t )
f1 p f
= N kp,1 (t ) −
f2 c

The scale factor on the L2 carrier phase is used to scale it to have the same frequency as in L1.

This destroys the integer ambiguity nature of phase on L2. The phase ionospheric

combination can be used to detect cycle slips. Figure 8.2 shows a phase ionospheric plot.

5.3.4.3 Single-Frequency Ionospheric-Free Code and Phase

Since the ionospheric effect disrupts the code and phase differently (thus the term "group

delay and phase advance"), it is possible to use this knowledge to eliminate common

ionospheric error:

Pkp, i (t ) + Φ kp, i (t )
Ξ kp, i (t ) =
2
= ρ kp (t ) − c dt k + c dt p + Tkp (t ) + N kp, i
5. 44
d k ,i , P (t ) d kp,i , P (t ) d ip, P (t ) ε i,P
+ + + +
2 2 2 2
d k ,i ,Φ (t ) d kp,i ,Φ (t ) d ip,Φ (t ) ε i ,Φ
+ + + +
2 2 2 2

with the reparameterized ambiguity

c TGD, i
N kp, i = N kp, i + β c 5. 45
2 fi 2

89
This equation can be applied to both L1 (β = 1) and L2 (β = α). However, clock errors do not

cancel out, but multipath and noises contribute at half of the pseudorange and carrier phase

p
values. The analysis process must be able to estimate the reparameterized ambiguity, N k .

90
6 Precise IGS Orbit and Satellite Clock

The International GPS Service (IGS) generates precise ephemerides for the satellites together

with by-products such as Earth orientation parameters (EOP) and GPS clock corrections. The

IGS service is built upon a global network of permanent tracking stations and provides

information and data products from computational centers to all GPS users through data

archive and exchange centers. In this chapter, the IGS analysis is introduced. The GPS orbital

information is normally given in the standard SP3 format which will be discussed in detail in

the following section. Format and description of satellite clock corrections will also be

covered.

6.1 IGS Orbital Analysis and Its Products

6.1.1 IGS Structure and Operation

Since its establishment, the IGS accomplishes its mission through the following components:

• Networks of tracking stations

• Global and Regional Data Centers

• Analysis and Associate Analysis Centers

• Analysis Coordinator

91
• Central Bureau

• International Governing Board

The tracking data are available at various Data Centers, the individual orbits

determined by the Analysis Centers at the Global Data Centers, and the official IGS orbits are

combined at the Central Bureau and the Global Data Centers (IGS, 1998). Table 6.1 provides

the current IGS components/structure, beginning with the IGS Operation Centers and the IGS

station network that rigorously apply IGS standards for station monument/hardware, data

quality, submission formats, and delivery delays.

6.1.2 Products

Since the IGS test operations started in June 1992, a continuous set of highly accurate

daily GPS orbital data and EOP have been available from individual processing centers. Since

November 1992, the IGS Analysis Center Coordinator has regularly compared the orbits of

the individual processing centers (Goad, 1993). The IGS product accuracy has improved from

approximately 1m (orbits) and 1mas (EOP) (Beutler, 1994) to about 5 cm (orbits) and about

0.1 to 0.2 mas (EOP) (Neilan et al., 1997). This improvement indicates that the IGS orbit is

becoming more accurate, stable, and reliable. Table 6.2 shows approximate availability and

accuracy of IGS products. The EOP are combined with those determined by means of satellite

and lunar laser observations as well as VLBI observations by the IERS (Beutler, 1994). After

the discontinuation of SA, the final satellite clock correction has accuracy 0.1 ns (Kouba,

2000).

92
Table 6. 1. The current IGS structure/components

Operation Centers/ IGS Station Network:


Global Data Centers:
CDDIS - Crustal Dynamics Data Information System, GSFC, NASA, USA
IGN - Institut Geographique National, France
SIO - Scripps Institute of Oceanography, Univ. of Cal., USA
Regional Data Centers:
AUSLIG - Australian Land Information Group, Australia
BKG - Bundesamt für Kartographie und Geodäsie, Germany
JPL - Jet Propulsion Laboratory, Caltech, USA
NGS - National Geodetic Survey, NOAA, USA
NRCan - Geodetic Survey, NRCan, Canada
Analysis Centers:
CODE - Center For Orbit Determination in Europe, Univ. Bern, Switz.
NRCan - Geodetic Survey, NRCan, Canada
ESA - European Space Operation Center, ESA, Germany
GFZ - GeoForschungsZentrum, Potsdam, Germany
JPL - Jet Propulsion Laboratory, Caltech, USA
NGS - National Geodetic Survey, NOAA, USA
SIO - Scripps Institute of Oceanography, Univ. of Cal., USA
Associated Analysis Centers (AAC):
USNO - AAC for Rapid Service, U.S. Naval Observatory, USA
JPL - Global Network AAC (GNAAC), Jet Propulsion Lab., USA
MIT - GNAAC, Massachusetts Institute of Technology, USA
NCL - GNAAC, University of New castle upon Tyne, UK
RNAACs – 15 Regional Network AACs in Europe, N.A., Asia and Australia
Analysis Center Coordinator : (T. Springer/CODE)
Central Bureau (CB): hosted by JPL, includes IGS CB Information System
(Director: R. Neilan)
International Governing Board: 15 members

Normally, the high-quality GPS data is online within one day and data products are

online within two weeks of observations. The IGS global network of permanent tracking

stations, each equipped with a GPS receiver, generates raw orbit and tracking data. The

Operational Data Centers, which directly contact the tracking sites, collect the raw receiver

data in Receiver INdependent EXchange format (RINEX) (Gurtner, 1997) and then forward

these data to the Regional or Global Data Centers. For efficiency and to reduce electronic

network traffic, the Regional Data Centers collect data from several Operational Data Centers

before transmitting them to the Global Data Centers. Data not used for global analyses are

archived and available online at the Regional Data Centers. The Global Data Centers archive

93
and provide online access to tracking data and data products which normally must be available

to users for at least 60 days (Kouba et al., 1998). The online data are employed by the

Analysis Centers to create a range of products which are then transmitted to the Global Data

Centers for public use. The IGS Central Bureau Information System, accessible on the

Internet, provides both IGS member organizations and the public with a gateway to all the IGS

global data and data product holdings along with other valuable information.

Table 6. 2 Approximate Availability and Accuracy of the IGS Products


Units: mas – milli-arc-second; ms - millisecond (Kouba et al., 1998) Note that the predicted
clock accuracy refers to the case of SA being active

IGS Products Availability Interval Accuracy


Ephemerides
Predicted Real Time 15 min 50 cm
Rapid 1-2 days 15 min 10 cm
GPS Final 10-12 days 15 min 5 cm
Satellites Clocks
Predicted Real Time 15 min 150 ns.
Rapid 1-2 days 15 min 0.5 ns.
Final 10-12 days 15 min 0.3 ns.
IGS Station
Weekly Solutions < 4 weeks 7 days 3-5 mm
Positions
Pole
Rapid 1-2 days 1 day 0.2mas
Final 10-12 days 1 day 0.1mas
Pole Rates
Rapid 1-2 days 1 day 0.4 mas/day
Earth Final 10-12 days 1 day 0.2 mas/day
Orientation UT1−UTC
Rapid 1-2 days 1 day 0.20ms
Final 10-12 days 1 day 0.05ms
Length of Day
Rapid 1-2 days 1 day 0.06ms/day
Final 10-12 days 1 day 0.03ms/day
Tropospheric Zenith Delay < 4 weeks 2 hours 0.4cm

Since January 1994, contributions from the seven current IGS Analysis Centers make

IGS official orbits possible and available to a user community. Other than supporting a

variety of government and commercial interests, the IGS also develops international GPS data

94
standards and specifications. With a multi-national affiliation, the IGS collects, archives, and

distributes GPS observation data sets with sufficient accuracy to satisfy the objectives of a

wide range of applications and experiments. IGS uses these data sets to generate data

products as follows:

§ High-quality orbits for all GPS satellites (estimated accuracy better than

5 cm (one sigma) for the final ephemerides)

§ Earth Rotation Parameters

§ Contributions to determine the tracking site coordinates and velocities in

International Terrestrial Reference Frame (ITRF), in close cooperation

with the International Earth Rotation Service (IERS)

§ Phase and pseudorange observations in daily RINEX files for each IGS

tracking site

§ GPS satellite and tracking station clock information

§ Ionospheric information

§ Tropospheric information

§ Other data products in support of geodetic and geophysical research

activities.

Highly accurate and reliable data and data products supplied by the IGS that meet the

demands of a wide range of applications and experimentation are available within two weeks

of observation. These data can be accessed through the Internet via the Information System

managed and maintained by the IGS Central Bureau (JPL), which is sponsored by the National

Aeronautics and Space Administration (NASA).

IGS near real time high-quality GPS data and data products provided by the IGS

global system of satellite tracking stations, Data Centers, and Analysis Centers meet the

objectives of a wide range of scientific and engineering applications, and research. Even

95
though improvements are frequently made, the current accuracy of various IGS products is

sufficient to support, improve, and extend current scientific objectives including:

• Realization of global accessibility to ITRF

• Improvement of ITRF

• Monitoring deformations of the solid Earth

• Monitoring Earth rotation

• Monitoring variations in the liquid Earth (sea level, ice-sheets, etc.)

• Scientific satellite orbit determinations

• Ionospheric monitoring

• Climatological research, eventually weather prediction

Precise GPS Satellite Ephemerides and Clock Information: To obtain the final orbits, the

combination of ephemerides can be accomplished via many means. The official IGS orbits

are produced by forming a weighted combination of the ephemerides submitted by each

individual analysis center (IGS, 1997). Each week the Analysis Center Coordinator provides

the precise orbits and clocks and a summary report file documenting the combination process.

Based upon IGS orbit comparison and combination, final GPS satellite ephemerides precision

is better than 5cm (one sigma) in each coordinate. Normally, a rapid solution can be

computed from 15 to 20 worldwide distributed stations after the end of the day using data

available at the time. The rapid solution is made available usually within 21 hours following

the observations with its estimated accuracy better than 50cm (one sigma) in each coordinate

component. Generally, this degradation over the final solution has minor impact on the

positioning accuracy for most GPS users. Approximate availability and accuracy of the IGS

products are listed in Table 6.2 (Kouba et al., 1998).

96
IGS has taken two actions to improve the consistency between the combined IGS

orbits and the combined IGS clocks. First is an improved clock weighting scheme using the

clock estimates from one AC as reference instead of the satellites without SA. Second is to

correct the AC clock, before the combination, based on the difference in the radial component

between the AC orbit and the IGS combined orbit. It is expected that all high quality IGS

products should be more reliable and at least as accurate as, if not more than, the solutions

obtained from each individual analysis center (Springer et al., 1998).

More attention has been given to the improved precision of the IGS combined orbit

prediction (IGP), LOD/UT combination and satellite clock combination. Comparing with the

IGS rapid solution (IGR), the IGP outlier detection can be performed (Kouba et al., 1998).

This considerably enhances the IGP reliability and consistency. The resulting ephemerides

and clocks are output to daily files in the SP3 format and these product files can be obtained

from the CBIS or any of the Global Data Centers. Table 6.3 shows the IGS and IGR

comparisons with the IERS Bulletin A for 1997.

Table 6. 3 Comparisons of IGS Rapid and IGS Final combined EOP with the IERS
Bulletin A for 1997 (units: mas – milli-arc-sec.; ms - millisec.) (Kouba et al., 1998)

IGS Final IGS Rapid


Comparison PMx PMy LOD UT PMx PMy LOD UT
(mas) (mas) (ms) (ms) (mas) (mas) (ms) (ms)
Mean .28 .15 .001 .015 .40 .26 −.004 .043
Standard
.07 .07 .026 .044 .24 .27 .034 .203
Deviation

It became clear that the IGS orbits and clocks were inconsistent at the 200 mm level

and improvement has been carried out by improving the clock weighting scheme and

correcting the AC clocks before the combination (Springer et al., 1998). The update and

97
combination strategies including orbit combination and evaluation statistics and remarks are

given in the IGS final summary file. The current IGS products are based on ITRF-97.

6.2 The SP3 Ephemeris

The currently used GPS satellite orbital format distributed by the IGS is the format

defined by the National Geodetic Survey (NGS), called SP3 format. SP3 is an ASCII

representation that includes the satellites’ position and clock corrections. ECF3 and EF18 are

binary counterparts to SP3, which are regarded as the second generation of orbital formats.

All formats have been carefully designed by taking many factors into consideration including

their use for GLONASS and geostationary satellites (Remondi, 1993). While the previous

generation could handle only 35 satellites, the second generation can accommodate up to 85

satellites (NGS, 2000; Remondi, 1993). Additionally, the header section of the second

generation has been developed to allow changes and insert new information, e.g., orbital

accuracy information for each satellite.

6.2.1 The SP3 GPS Orbital Format and Data Accuracy

Following the NGS study, it appeared that for all application purposes the velocity data does

not need to be distributed, since it can be calculated to an accuracy about 0.004 mm/s from the

positional data (Remondi, 1993). However, the velocity data can optionally be included in the

SP3 format. NGS provides programs for users to recover velocity information as well as

translate one format to another. The SP3 format is given to 1 mm and 1 ps (pico-second). For

each information line for a given satellite, the flag notation “ P” for SP1, SP2, and SP3 refers

to position-only, and “ V” for SP3 indicates velocity. Both position and velocity are required

for the velocity flag. The velocity data has units of decimeters/s with an accuracy of 10−4

98
mm/s. The rate of change of clock correction at the last column of the velocity line has units

of 10−4 µs/s with precision of 1 ps/s. See appendix for the SP3 format.

6.2.2 Precise Satellite Clock Information

The satellite clock information is given in the SP3 data file in units of microseconds. It is

given at the same epochs for which the satellite positions are given. Its precision information

is discussed above: 1 ps for the clock and 1 ps/s for the rate of change of clock correction.

6.3 Lagrange Interpolation

GPS ephemeris and clock is given at a nominal epoch. The interpolation is needed to obtain

the satellite position and clock correction at the transmission epoch. Lagrange is the Newton

form of interpolating polynomials. Lagrange is probably the most convenient and efficient

method and has several advantages (Cheney and Kincaid, 1994, page 134). Lagrange can be

applied to the GPS orbit interpolation as utilized by Remondi (1989). We can also apply

Lagrange to the satellite clock interpolation as well.

Having a set of fixed nodes x1 , x 2 , K , x n , the system (called cardinal function in

interpolation theory) will have polynomials of degree (n − 1) . Such system has n special

polynomials denoted by l1 , l 2 , K , l n with the following property

0 if i ≠ j
li ( x j ) =  6. 1
1 if i = j

The Lagrange formula, a linear combination of polynomial li , can then be used to interpolate

any function f:

99
n
p( x) = ∑ li ( x) f i ( xi ) 6. 2
i =1

p(x) is the interpolating polynomial of degree (n − 1) for the function f. p(x) yields

f ( x j ) at x j . The formula for li , the product of (n − 1) linear factors, is given as follows:

n  x−x 
l i ( x) = ∏  
j
1≤ i ≤ n. 6. 3

j ≠i  x i − x j


j =1

li is therefore a polynomial of degree (n − 1) and having the required property given in

equation 6.1. Note that the denominators of li are just numbers.

For equal epoch intervals ephemeris and for a given degree of polynomial one can

pre-compute the denominators by standardizing the fixed nodes, e.g., 1, 2, …, n. This will

save computation time and resources. Table 6.4 provides pre-computed denominators of li

for the standardized fixed nodes.

Table 6. 4 Pre-computed denominators of li for the standardized fixed nodes

Number of
standardized Denominators of li
fixed nodes (n): i = 1, 2, …, n
1, 2, …, n
3 2 -1 2

5 24 -6 4 -6 24

7 720 -120 48 -36 48 -120 720

9 40320 -5040 1440 -720 576 -720 1440 -5040 40320


3628800 -362880 80640 -30240 17280 -14400
11
17280 -30240 80640 -362880 3628800

100
7 Mathematical Implementations

This chapter provides mathematical techniques and implementations utilized in PPP analysis.

These include cycle slip detection and removal and mathematical consideration associated

with Kalman filter implementation for PPP. In addition, the computation flow and software

components are also given. Information from previous chapters is used as fundamental

principles of the software development.

7.1 Dilution of Precision

The term dilution of precision (DOP) has been accepted to characterize the effect of

the geometric satellite distribution on the accuracy of the navigation solution. The DOP

factors include vertical dilution of precision (VDOP), horizontal dilution of precision (HDOP),

positional dilution of precision (PDOP), time dilution of precision (TDOP), and geometric

dilution of precision (GDOP). The DOP expressions are given below (Langely, 1999):

101
σh
VDOP =
σ
σ n2 + σ e2
HDOP =
σ
σ n2 + σ e2 + σ h2
PDOP = 7. 1
σ
σt
TDOP =
σ
σ n2 + σ e2 + σ h2 + σ t2 c 2
GDOP =
σ

where σ is usually taken to be equal to the total user equivalent range error (UERE) and

σ n2 , σ e2 , σ h2 , and σ t2 are the corresponding variances of position in Northing, Easting, Height

(Up), and Time. The DOPs vary from epoch to epoch according to the change of satellite

geometry.

7.2 Cycle Slip Detection and Removal

The carrier phase measurements can be continued only when the receiver has the ability to

maintain lock on the incoming GPS signal, which is directly related with the integer number

of wavelengths (called ambiguity) referenced to the first phase measurement when the

receiver starts to lock on the signal. But in some situations, the receiver loses lock of the

phase lock loop which is due to the poor reception, very fast acceleration changes, or

shadowing of satellite signals by obstacles in their path. Such situations cause discontinuities

in the phase measurements, called "cycle slips," triggering a sudden jump of the phase.

102
7.2.1 Multipath

Multipath varies greatly depending upon a variety of factors, for example, the receiver-

satellite-reflector geometry and the strength and the delay of the reflected signal compared to

the line-of-sight signal. Therefore, multipath is considered one of the major limitations

imposed on the accuracy of the observables. Multipath distorts the C/A-code and P-code

modulations as well as the carrier phase observation. Multipath affects pseudoranges much

more than phases, therefore phases provide more precise solutions. The multipath expressions

applied to GPS/GLONASS are given in equations 7.2 and 7.3 (Li, 1995),

Pkp,1 + (2 β − 1)Φ kp,1 − 2 βΦ kp, 2 = (2 β − 1)λ1p N kp,1 − 2 βλ2p N kp, 2


+ {d1, P }+ (2 β − 1){d1,Φ }− 2 β {d 2,Φ }
7. 2

Pkp, 2 + 2αβΦ kp,1 − (1 + 2αβ )Φ kp, 2 = 2αβλ1p N kp,1 − (1 + 2αβ )λ2p N kp, 2
+ {d 2, P }+ 2αβ {d 1,Φ }− (1 + 2αβ ){d 2,Φ }
7. 3

where

2
 f1 p 
α ≡  p
f 
 2 

1
β ≡
1−α

Φ k = measured carrier phase scaled to distance (meters)


p

{d } = d
i,Φ k , i, Φ
(t ) + d kp, i , Φ (t ) + d ip, Φ (t ) + ε k , i , Φ for carrier phase, i = 1, 2

{d } = d
i, P k ,i, P
(t ) + d kp, i , P (t ) + d ip, P (t ) + ε k , i , P for pseudorange, i = 1, 2.

103
Assuming there are no cycle slips and a specific satellite is being considered, the first two

terms on the right side of equations (7.2) and (7.3) should provide a constant straight line. The

rest of the terms can be considered noises from multipath and hardware delays. Accordingly,

a plot using the terms on the left side of both equations (i.e., the observations) can be used to

identify such noises. More importantly, cycle slips may be detectable to a certain extent from

such plots.

Figure 7. 1 Multipath on P1 and P2 for PRN29 [NJIT, DOY137(2000)]

Multipath plots on P1 (equation (7.2)) and P2 (7.3) are illustrated in Figure 7.1,

together with the elevation angle. Multipath appears to be greater at low elevation angles.

The plots of equations (7.2) and (7.3) can be used in the preprocessing step to visualize quality

of the data as well as to detect cycle slips.

7.2.2 Widelane

Widelane is a linear combination of carrier phase that increases the effective GPS signal wave

length. Widelane ambiguity can be computed from pseudorange and carrier phase

observation.

104
f1 P1p + f 2 P2p
N wp = ϕ wp − 7. 4
( f1 + f 2 ) λ w

where

ϕ wp = ϕ1p − ϕ 2p 7. 5

and

c
λw = ≅ 86.2 cm 7. 6
fw

f w = f1 − f 2 7. 7

Widelane is a useful linear combination for the purpose of ambiguity estimation in

differential GPS (Teunissen, 1997). Theoretically, widelane ambiguity should be constant.

Due to spatial changes of the satellite signal's path causing multipath, the widelane ambiguity

function has a small variation particularly at low satellite elevation angles.

7.2.3 Cycle Slips

Attempts can be made to detect and correct the change in phase ambiguity. If a cycle-slip is

found and fixed, no further action is needed. But in the case of loss of lock the respective

ambiguity must be re-estimated.

7.2.3.1 Between Satellite Differences

A cycle slip causes a sudden jump in the phase observation by an integer number of cycles and

all observations obtained after the cycle slip are shifted by the same integer amount. One way

to find a cycle slip is to plot phase observations versus time which will show a step function or

105
individual outliers. Cycle slips can be any size ranging from one to millions of cycles. The

plot should have the capability to be viewed at different scales (or different aspect ratios).

Typically, steep slopes are seen from the plots of undifferenced observations as a result of

receiver clock error. In such cases, small slips may not be visible.

For single stations, a useful function for slip detection is the between-satellite single-

difference (one receiver and two satellites) of carrier phases as presented in equation 7.8.

ϕ k (t ) ≡ ϕ k (t ) − ϕ k (t ) .
pq p q
7. 8

Notice that equation 7.8 is free of receiver clock error. Slips are removed by adding integer

values to restore continuity of the carrier function. However, visual cycle slip fixing is not

easy because (7.8 ) is strongly dependent on time.

Carrier phase OMC (Observed Minus Computed) between satellites shows less of a

time dependency. Receiver clock error is eliminated. When satellite clock correction is not

applied, a slope might still be visible due to high clock drift. See Figure 7.2. The plots use

observations after SA had been turned off. A high variation would be seen with SA-on

observations. The satellite clock correction removes most of the slope in the OMC plot as

seen in Figure 7.3. Note that slips do not necessarily produce integer steps in the ionosphere-

free function (Leick, 1995, page 356). Single-frequency users are limited, of course, to L1

OMCs only. To detect cycle slips, the Kalman filter (Mertikas and Rizos, 1997) may be

applied to the OMC.

106
Figure 7. 2 Phase OMC between satellites without satellite clock correction applied.

L1 (top), L2 (middle), and ionosphere-free (bottom). The base satellite is PRN25.

[WES2, DOY138(2000)]

107
Figure 7.3 Phase OMC between satellites with satellite clock correction applied.

L1 (top), L2 (middle), and ionosphere-free (bottom). The base satellite is PRN25.

[WES2, DOY138(2000)]

108
Figure 7. 4 Phase OMC between satellite ionosphere-free PRN21-PRN25 without (top)

and with (middle) satellite clock correction applied, and satellite clock

correction difference (bottom). [WES2, DOY138(2000)]

Figure 7.4 shows the ionosphere-free phase OMC without and with satellite clock

correction applied and satellite clock correction for the difference PRN21-PRN25. The

IGS(JPL) produced satellite clock correction has the same but mirrored variation as seen in the

OMC when satellite clock correction is not applied (compare top and bottom figures). The

109
reversal seen after epoch 550 in the middle plot reveals possible errors in the satellite clock

correction. Prior to the reversal the clock corrections are not available for several epochs.

7.2.3.2 Undifferenced Observation Cycle Slip Detection and Fixing

Undifferenced cycle slip detection and fixing as discussed in Blewitt (1990) has been

implemented in this study. Since cycle slips can occur concurrently and differently on L1 and

L2, the slip on each signal must be independently detectable. The undifferenced cycle slip

detection involves linear combinations of the observations: the widelane combination and the

ionospheric combination. Let N1 and N2 be ambiguities for L1 and L2, respectively.

7.2.3.2.1 Cycle Slip Detection in the Widelane Combination

The widelane ambiguity is written as:

f1 P1 + f 2 P2
Nw = ϕw − . 7. 9
( f1 + f 2 ) λ w

For every observation epoch, the widelane ambiguity is evaluated. Setting a priori rms of half

widelane cycles, the sequential recursive algorithm for the widelane ambiguity and its

variance is given as

Nw i
= Nw i −1
+
1
i
[N w,i − N w i −1
] 7. 10

σ i2 = σ i2−1 +
1
i
{(
N w,i − N w i −1
)2 − σ i2−1 } 7. 11

where N w is the mean value of Nw, σi is the standard deviation of Nw , and i refers to the

current data epoch being evaluated. The subsequent epoch is required such that N w,i +1 is

110
within 4σi of the running mean N w i . Otherwise it is assumed that a cycle slip has occurred

and the respective number of widelane cycle slips is recorded. However, consecutive slips

with equal widelane slips but negative sign are later treated as outliners in the cycle slip fixing

stage.

7.2.3.2.2 Cycle Slip Detection in the Ionospheric Combination

The idea of ionospheric slip detection is used in the very unlikely case that the slip in L1

equals the slip in L2, which the widelane slip detection is unable to detect. The ionospheric

combination for carrier phase observation is

ϕ ϕ 
Φ iono = c  1 − 2 
 f1 f 2 
= δI + λ1 N1 − λ 2 N 2 7. 12
= δI + λ1 ( N1 − N 2 ) + (λ1 − λ2 ) N 2
= δI + λ1 N w − λ I N 2

f 2 
where δI =  12 − 1 I P1 and IP1 is the ionospheric delay on P1-code. λ I = (λ1 − λ 2 ) is the
 
 f2 

ionospheric wavelength.

The pseudorange ionospheric combination is

Piono = P2 − P1
. 7. 13
= δI

Since the pseudorange does not have integer cycle discontinuity, it is possible to construct a

polynomial fit Q to Piono and subtract it from Φ iono ( i.e. Φ iono − Q ), then look for

discontinuities. Blewitt (1990) suggested an empirical formula to compute the degree of

polynomial fit as

111
 N 
m = min  + 1, 6 7. 14
100 

where N is the number of observations in the data. It should be noted that this simple fit is

only used for discontinuity detection, not for the value of cycle slip. The algorithm is given as

( LI i − Qi ) − ( LI i −1 − Qi −1 ) > k cycles 7. 15

( LI i +1 − Qi +1 ) − ( LI i − Qi ) < 1 cycles 7. 16

where epoch i is the first good data point after the occurrence of a cycle slip if both of the

above conditions are met. The default value of k is set to 6 ionospheric cycles (6 x 5.4 cm =

32.4 cm), but can be set to a more appropriate value depending on the ionospheric conditions.

The reasons for using high tolerance k value are because

(1) receivers at high latitudes often have large phase variations due to ionospheric activity.

This should not be confused with cycle slips.

(2) there are only slim chances of having equal cycle slips in L1 and L2 of less than 6

ionospheric cycles.

Equal cycle slips are detected by the ionospheric combination, the respective widelane

cycle slips are therefore set to zero.

7.2.3.2.3 Fixing Cycle Slip of Undifferenced Observation

Let ∆N1 and ∆N2 be the number of cycle slips for L1 and L2, respectively. The number of

widelane slips is: ∆Nw = ∆N1 − ∆N2 . The cycle slip can be proceeded by a polynomial fit to

Φ iono just before the slip occurrence and then extrapolated to the slip epoch or after the slip

epoch if in case of a data gap. This is denoted as Φ iono, estimate . The value of real Φ iono at

112
the slip epoch is also calculated, Φ iono, at _ slip . The difference is

∆Φ iono = Φ iono, at _ slip − Φ iono, estimate . We can apply the knowledge from Φ iono :

∆Φ iono = λ1∆N w − λ I ∆N 2 . 7. 17

∆Nw is obtained from the widelane slip detection, or zero in case of equal slip detected by the

ionospheric algorithm. ∆N2 can therefore be computed from the above equation. Since ∆Nw

= ∆N1 − ∆N2 , now the number of cycle slips ∆N1 in L1 is very obvious. Having ∆N1 and ∆N2

values, cycle slips in the phase data can be fixed by adding (or subtracting) the integer cycles

to all the respective subsequent data points.

7.3 Kalman Filter

The Kalman filter is an optimal estimation technique that minimizes the estimation error in a

well defined statistical sense (Gelb, 1974). As a linear filter using a recursive algorithm which

processes measurement information sequentially in time, the Kalman filter involves two main

steps: filtering and prediction. Filtering is the estimation of the state vector at the current

epoch based on all previous measurement information. Prediction involves the estimation of

the state vector x at a future time.

The Kalman filter system state vector (dynamic model) which evolves with time can

be written as

x k +1 = Φ k x k + wk 7. 18

corresponding to the measurement vector (measurement model)

z k = H k xk + vk 7. 19

113
where wk ~ N k (0, Qk ) and v k ~ N k (0, Rk ) are the system and measurement noises which

are mutually uncorrelated vectors. Subscript k refers to the epoch of time. Φ is the transition

matrix. H is the measurement connection matrix. Elements of H are the partial derivatives of

the predicted measurements with respect to each stage and must be computed for every epoch.

The respective Kalman filtering algorithm involves Kalman Gain (K), covariance update ( Pk )

and prediction ( Pk−+1 ), in the time update and measurement update steps, as shown in Figure

7.5.

114
− −
Initial Estimates of xˆ k and Pk

Measurement Update (Filter or correct)

(a) Calculate Kalman Gain

K = Pk− H kT ( H k Pk− H kT + Rk ) −1

(b) Using measurement to update estimate

xˆ k = xˆ k− + K ( z k − H k xˆ k− )

(c) Update the error covariance

Pk = (I − K k H k )Pk−

Time Update (Predict)

(a) Project the state ahead

x k−+1 = Φ k xˆ k

(b) Project the covariance ahead

Pk−+1 = Φ k Pk Φ Tk + Qk

Figure 7. 5 The Kalman filter computation recursive scheme

115
7.3.1 Extended Kalman Filter

The extended Kalman filter (EKF) is an extension of the standard Kalman filter (SKF). EKF

is used when the process to be estimated and/or the measurement relationship to the process is

non-linear (Welch and Bishop, 1997). EKF simply applies SKF through the linearization

(through approximation of Taylor series) around the previous state estimate.

The EKF provides a simple but efficient algorithm to process a nonlinear system. In

this work, we deal with a nonlinear measurement system and a linear dynamic model.

The state vector system which evolves with time can be written as

x k +1 = Φ k x k + wk 7. 20

corresponding to the measurement vector

z k = h( x k ) + v k 7. 21

The respective governing equations that linearize an estimate about 7.20 and 7. 21 are

x k +1 = Φ k xˆ k + wk 7. 22

zk = ~
z k + H ( xk − ~
xk ) + vk 7. 23

H is a partial derivative of measurement equation h(o) with respect to the estimated vector x,

around the approximated state vector. The EKF computation recursive scheme is given in

Figure 7.6.

116
Initial Estimates of xˆ k− and Pk−

Measurement Update (Filter or correct)

a) Calculate Kalman Gain

K = Pk− H kT ( H k Pk− H kT + Rk ) −1

b) Using measurement to update estimate

xˆ k = xˆ k− + K ( z k − h( xˆ k− ))

c) Update the error covariance

Pk = ( I − K k H k )Pk−

Time Update (Predict)

a) Project the state ahead

x k−+1 = Φ k xˆ k

b) Project the covariance ahead

Pk−+1 = Φ k Pk Φ Tk + Qk

Figure 7. 6 The EKF computation recursive scheme

117
7.3.2 Discrete Gauss-Markov Process

A Markov process is a random process that allows users to link the process with simple filters.

A fundamental model is first-order Markov if the probability distribution for the process ( x k )

depends only on the value at one point immediately in the past ( x k −1 ) (Gelb, 1974, page 42).

The differential equation for continuous first-order Markov process x(t) is given as

1
x& (t ) = − x(t ) + w(t ) 7. 24
τ

where τ is the correlation time and w is white noise. If the condition is added that the

probability density function of w and therefore x also are Gaussian, the given process is a

Gauss-Markov process. A discrete version (or a first-order difference equation) of a Gauss-

Markov process can be written as

xk +1 = κ xk + wk 7. 25

where

T
κ= eτ 7. 26

T is the data interval. Let β = (1 / τ ) , β is called the dampening coefficient. A larger β value

yields a shorter correlation length. This permits a large variable change from one epoch to the

next. On the other hand, a small β value describes high correlation in the following epochs,

and thus allows only a small variation (Mertikas and Rizos, 1997). The associated covariance

in discrete time for the Gaussian white noise w is

q
cov(w) = [1 − e −2Tβ ] . 7. 27

118
q is the variance of the process noise. If τ equals zero, the model becomes a pure white noise

model without correlation time which implies κ = 0 . But if the correlation time τ

approaches infinity [ κ = 1 ], the process is called pure random walk (Gelb, 1974 pages 43 and

79).

Station coordinates are usually modeled as pure random walk, so are reparameterized

ambiguities. Troposphere and receiver clock can be modeled as either white noise or random

walk (Tralli et al., 1990).

7.3.3 PPP Implementation

Mathematical issues for PPP implementation are given. These include partial derivatives for

EKF, receiver clock estimation, ambiguity estimation, and observation weighting scheme.

7.3.3.1 Partial Derivatives

The submatrix H for satellite p at epoch i for the state vector of position x, y, z, ambiguity,

receiver clock, and tropospheric effect can be written as.

 xi− − X ip y i− − Yi p z i− − Z ip 
H ip (Φ ) = 1 1 TMF 
 ρ i ρ ip ρ ip
p

 xi− − X ip y i− − Yi p z i− − Z ip 
H ip ( P) = 0 1 TMF  7. 28
 ρ i ρ ip ρ ip
p

ρ ip = ( xi− − X ip ) 2 + ( y i− − Yi p ) 2 + ( z i− − Z ip ) 2

p p p
where ( X i , Yi , Z i ) are the coordinates of satellite p at the transmission time and

− − −
( xi , y i , z i ) are the estimated or projected receiver coordinates. For the carrier phase

observation Φ (scale ϕ to distance), the partial derivative for the reparameterized ambiguity of

119
the respective satellite is 1. The partial derivative for zenith troposphere is the tropospheric

mapping function (TMP) (e.g., NMF). Rows of H for dual-frequency analysis are twice that

for single-frequency.

7.3.3.2 Receiver Clock Estimation

Receiver clock can be estimated in EKF using a Gauss-Markov process. Zumberge et al.

(1997b) suggested that the receiver clock can be modeled as white noise. However, it is better

to model it as random walk. Since we have pseudorange observations which include the clock

term, an approximate ratio of the change in the clock from the current epoch to the next in the

EKF dynamic model can be directly computed.

Ri = Pi p − ( ρ ip − cdtip )
7. 29
Ri +1 = Pi +p1 − ( ρ ip+1 − cdtip+1 )

Ri +1
Φ clk = 7. 30
Ri

Because many satellites are observed at the same time, it is additionally better to average out

the random error by averaging the ratio of all available satellites. The PPP solution obtained

from the random walk process with a priori information of the ratio of receiver clock shows a

slight solution improvement than that using the white noise process.

7.3.3.3 Ambiguity Estimation

The ambiguity should be constant as long as the receiver is locked to the signal. The

ambiguity number changes when the receiver loses lock. Unlike the double differenced

carrier phase, the undifferenced PPP analysis estimates only real numbers of ambiguities, or

120
"floated solutions." In fact, PPP ambiguities are reparameterized ambiguities which are

collections of ambiguities of L1 and L2, with carrier phase ionosphere-free scaling factors.
)
The reparameterized ambiguities N are being estimated every epoch, as a pure random walk

process. The dynamic model is in the form

) )
N ip+1 = N ip + w 7. 31

where w is Gaussian noise.

7.3.3.4 Observation Weighting Schemes

Recent scientific GPS research has started to pay more attention to observation weighting

methods (Vermeer 1997; Teunissen, 1998; Collins and Langley 1999; Hartinger and Brunner

1999). Because different weight methods yield different solutions particularly in the height

component, the observation must be given appropriate weight. As alternatives to constant

weighting, normally used in GPS analysis, other possible weighting schemes may be used.

For example:

a) Exponential weighting schemes that weight corresponding observations observed from

near horizon satellites is lowered (Euler and Goad, 1991)

b) Weighting that reflects receiver generated signal-to-noise ratio (SNR) values (Collins and

Langley, 1999; Hartinger and Brunner, 1999). Normally, SNR represents the carrier-to-

noise-power-density ratio (C/N0) which varies with the elevation of the arriving signal.

Langley (1997) derived phase variance (m) using C/N0 (dB-Hz) values as follows

2
λ 
σ Li
2
= B  i  10 −0.1 C / N0 , i = 1, 2 7. 32
 2π 

121
where B is the carrier tracking loop bandwidth (Hz). It should be noted that some receiver

manufacturers do not provide either SNR or C/N0, and some provided are in an arbitrary

format.

c) Weighting as a cosecant function of the satellite elevation angle (Vermeer, 1997; Collins

and Langley, 1999). This is because the amount of signal noise increases towards the

horizon, similar to the tropospheric error which has a cosecant shape, according to various

models of the tropospheric mapping function (e.g., Marini, Chao, Davis, and Herring

mapping functions).

d) Weighting as square of a cosecant function of the satellite elevation angle (Vermeer,

1997; Hartinger and Brunner, 1999). This is from the fact that GPS residuals reveal a

more swiftly increasing noise level for low elevation angles.

e) Combination weighting using C/N0 information together with satellite elevation

knowledge.

Figure 7. 7 Potential weighting functions, comparison between cosecant and cosecant-

squared

122
Hartinger and Brunner (1999) used the SIGMA-ε model where the phase variances

are computed using C/N0 values and thus observation weight directly echoes signal quality.

Their experimental results show that baseline rms of the SIGMA-ε model is much less than

that of equal-weighting, especially at a low elevation cutoff angle.

Collins and Langley (1999) reported that, in the presence of multipath, the cosecant

and SNR weighting schemes yield a great improvement over the equal-weighting scheme.

Moreover, according to the scaling effect of the a posteriori variance factor, the cosecant and

SNR schemes are almost numerically equivalent.

The amount of observation noise increases and exhibits presence of multipath which

mostly occurs in signals from low satellite elevation angles. It should be more appropriate to

apply a step function using a combination of uniform weight for high elevation angle

observation and lower weight at low elevation angle. Deweighting observations at high

elevation angles will lose valuable information. The step function variance may be given as

σ 2 E >α
σ (E ) =  2
2
7. 33
σ cosec 2 ( E ) E <α

or

σ 2 cosec( E ) E >α


σ 2 (E ) =  2 7. 34
σ cosec 2 ( E ) E <α

where α is the elevation threshold angle.

123
Figure 7. 8 Potential step function weightings, comparison between equations 7. 33 (line)

and 7. 34 (dotted line)

With PPP experiments, the solutions obtained from the step function weighting

scheme appear to be better than that from continuous varying weighting schemes.

7.3.4 Computational Flow and Software Components

This section provides PPP analytical detail starting from observation in RINEX format until

the result analysis. The summarized PPP algorithm used in the study is given in Figure 7.9.

Most of the program components used in this study have been written in Mathcad software.

Due to the creative design, the program is an interactive graphic-aided tool suited for GPS

research experiments involving PPP.

The Batch.exe program transfers the RINEX observation and the SP3 ephemeris into

column format which outputs are in the same satellite PRN sequence. Output files are written

on the computer disk. The columnized observations of carrier phase and pseudorange are used

in the cycle slip detection and removal step. After completion of the cycle slip analysis, the

124
data is trimmed to 30-second interval if necessary and the output file written to disk. The user

has the ability to visually examine the data quality via widelane, phase ionospheric

combination, and multipath plots, before and after running the cycle slip analysis. Comments

are noted for data deletions (DEL) or resetting the weight matrix Pk− (WT) for specific

epochs and satellites. The next step involves orbit and clock interpolation. The pseudorange

OMC computation is also carried out. After undesired data is deleted, the output is written to

a file. Next, before running the EKF, the relevant parameters and options can be set.

Resetting the weight matrix WT enters here, just before running EKF.

125
Observation in RINEX format
Satellite Ephemeris in SP3 format

Batch.exe

Observation in column format per Orbit & satellite clock in


satellite column format per satellite

Hour jump corrections (if any)


Orbit and clock interpolation to the
transmission epoch using Lagrange,
Cycle slip detection & removal and compute pseudorange OMC
Take a note tables DEL & WT

Pseudorange quality control


& data deletion (table DEL)

Update Update Extended Kalman Filter (EKF) with


table DEL table WT Gauss-Markov dynamic model

Accounted for Solid Earth Tides

Figure 7. 9 PPP Algorithm

126
8 Numerical Study and Results

Data sets were arbitrary selected from CORS sites, including data sets before and after

discontinuation of SA. The numerical detail is given. An experimental example in graphical

format allows the reader to follow visually. Numerical results and discussion will be

presented. Various components which affected the results are discussed, including zenith

troposphere (estimated or approximated via Saastamoinen model), impact of satellite antenna

offsets, impact of relativity, impact of a single cycle slip, the use of a priori tropospheric

information. Additionally, kinematic positioning experiments for dual and single-frequency

are also presented.

8.1 Data Sets

Observation data from four stations are used in this study. Data sets were downloaded from

the CORS network database, except the USNO station where data sets were downloaded from

the CDDIS (2000b) database. Each station apparatus is given in Table 8.1. All data sets used

in this study were limited to the first 6 hours of the day (0-6 GPS time).

The Antenna Reference Point (ARP) published coordinates are given in Table 8.2.

All coordinates, except the USNO station, are ITRF-96 positions at epoch 1997.0 which is

made available from NGS, CORS database. Coordinates of USNO station were obtained from

127
the station's latest log file. Table 8.3 shows data sets day of the year (DOY) for year 2000

that were used for this study. JPL's high-rate (30-second intervals) precise ephemerides were

used (JPL, 2000c).

Table 8. 1 Station hardware

Station Name Antenna Frequency


Receiver Type
(station ID, location) Type Standard
Westford Rogue AOAD/M_T
H-Maser
(WES2, Westford, MA) SNR-8000 Choke-ring

New Jersey Inst. of Tech. Leica LEIAT303


Internal
(NJIT, Newark, NJ) SR9500 Choke-ring

The Surveyors Exchange Anchorage Leica LEIAT504


Internal
(TSEA, Anchorage, AK) CRS1000 Choke-ring

US Navy Observatory Rogue AOA AOAD/M_T Steered


(USNO, Washington DC) SNR-12 ACT Choke-ring H-Maser

Table 8. 2 Published coordinates of ARP

Station
X Y Z Note
ID
Published by IERS
WES2 1492233.388 -4458089.474 4296045.997
in Jan., 1998

Computed in Feb. 1999


NJIT 1319479.773 -4656039.407 4140717.359
using 26 days of data

Computed in July, 1999


TSEA -2666154.987 -1545828.568 5565470.306
using 21 days of data

Data Aug-Sept 1997


USNO 1112189.903 -4842955.032 3985352.238
Epoch 1997.653

128
Table 8. 3 RINEX data sets
Legends: Y: Yes, this data was used, N: Not Used, NA: Not Available

DOY 002 065 100 122 128 136 137 138


WES2 Y N N Y Y Y Y Y
NJIT Y N N Y Y Y Y Y
TSEA NA Y NA Y N Y N Y
USNO NA NA Y Y Y N Y Y

8.2 A Priori Kalman Filter Settings

Kalman Filter settings involve the variance matrix Q of parameter noise, and the variance

matrix R of observation noise. The initial parameter variance Pk− is set at the beginning.

Since the Kalman algorithm employs both the measurement update and time update, the

parameter variance matrix P is adjusted to reflect the quality of the estimated parameter. The

variance Q affects Pk− in the time update stage, whereas in the measurement update Pk− , R,

and H influence the Kalman Gain K which in turn affects the variance P. Therefore, the

Kalman filter performance depends upon setting these quantities to the appropriate values that

reflect the real quality of the respective parameters.

It should be noted that the settings given here apply to a 30-second data interval.

These values are typically assumed in GPS analysis. The values for the variance matrix Q are:

3600 sec/ hr
σ x = 0.0003 m Coordinates. Multiplying this with will result in
30 sec

standard deviation in units m/hr.

σ clk = 30 m Receiver clock

129
σN = 0m Ambiguities. We know that ambiguities are constants, thus their noise

variances are set to echo this knowledge.

0.01m
σT = Zenith troposphere
3600 sec

The variance matrix R should be set such that it reflects the quality of the observation. We use

the following values:

σ Ph = 0.02 m For carrier phase ionosphere-free combination

σ Pr = 0.20 m For pseudorange ionosphere-free combination

Initial setting for the unknown parameters uses large values reflecting the accuracy of the

initial estimated unknown parameters. We use the following values in matrix P:

σ coor _ init = 1 m Coordinates

σ clk _ init = 500 m Clock

σ N _ init = 1010 m Ambiguities

σ T _ init = 0.1 m Zenith troposphere

8.3 Analysis Example

This section provides an example of a PPP analysis. One solution is documented from the

beginning to the end. The PPP algorithm used in the study is summarized in Figure 7.9. As

130
part of the analysis, graphical images in each analytical step are given. Station WES2,

DOY2(2000) is used for this illustration.

8.3.1 Widelane

Widelane is used in the cycle slip detection and fixing step. Discontinuities or jumps in

widelane indicate cycle slips. However, one should not confuse cycle slips with noise. Most

of the widelane variation comes from pseudoranges which are less precise than carrier phases.

No slip is found in Figure 8.1.

Figure 8. 1 Widelane before and after cycle slip detection (top) and the

difference(bottom) for PRN5 [WES2, DOY2(2000)]

131
8.3.2 Ionospheric Carrier Phase

The plot of ionospheric carrier phase can be used to identify equal cycle slips in L1 and L2

which the widelane does not show. However, there is a hidden problem in the ionospheric

carrier phase combination for certain slip combinations. For example, a pair of a nine-cycle

slip in L1 and a seven-cycle slip in L2 produces

f1 p
ϕ Ip = ϕ1p − ϕ
f2 2
 1575.42 
= 9−  7 = 0.017
 1227.6 

For such a combination, the slips cannot be detected because no ionospheric slip is found.

Therefore, the widelane is used to detect this cycle slip family. A plot of ionospheric carrier

phase is shown in Figure 8.2, which indicates no slip.

Figure 8. 2 Ionospheric carrier phase for PRN5 [WES2, DOY2(2000)]

8.3.3 Multipath

Multipath plots are used to visualize the quality of P1 and P2 observations. Large size cycle

slips in carrier phase may be also detected from the plots. Multipath plots are intended to

132
provide general information about data quality other than the cycle slip detection. From the

plot in Figure 8.3, P2 appears to have larger multipath than P1.

Figure 8. 3 Multipath on P1 and P2, for PRN5 [WES2, DOY2( 2000)]

8.3.4 OMC for P1

The computed OMC values take the satellite clock correction, approximate tropospheric

effect, and relativistic effect into account. The topocentric distance is obtained from

approximate receiver coordinates, the satellite position at the transmission epoch using

Lagrange orbit interpolation, and considering Earth rotation during single travel time. Since

all relevant quantities but the receiver clock error are taken into the calculation, the value from

P1 OMC is mostly receiver clock error. From Figure 8.4, other terms hidden in the graph are

133
ionospheric variation, multipath and random noise. Systematic variation within the band seen

in Figure 8.4 is mostly due to the ionosphere.

Figure 8. 4 OMC for P1 for PRN5 (top) and all satellites (bottom) [WES2, DOY2(2000)]

8.3.5 OMC for P2

In Figure 8.5, the OMC plot for P2 is similar to that for PR1. The graph reveals a similar

trend caused by the receiver clock, but shows greater ionosphere patterns. Since the

ionospheric effect is an inverse function of the signal's frequency, the ionospheric effect on P2

is much greater than on P1. The plot shows such a case. Station WES2 uses the Rogue SNR-

8000 receiver which utilizes dual-frequency P-codeless (cross-correlation) technology.

134
Figure 8. 5 OMC for P2 for PRN5 (top) and all satellites (bottom) [WES2, DOY2(2000)]

8.3.6 OMC Pseudorange Ionosphere-free

The linear pseudorange ionosphere-free combination equation (5.40) is used to calculate the

observed quantity. Although we are in the ascending phase (near the peak) of solar activity 23

which complicates the ionospheric effect on GPS signals, the ionospheric error is eliminated

through the ionosphere-free L1-L2 combination at least to the first order. Since the

ionosphere term has been taken out, the graph for all satellites seems to be more compact

revealing a more precise uniform trend of the receiver clock.

135
Figure 8. 6 OMC for pseudorange ionosphere-free for PRN5 (top) and all satellites

(bottom) [WES2, DOY2(2000)]

8.3.7 Pseudorange OMC Difference Between Satellites

Pseudorange OMC difference between satellites removes the receiver clock error, which is

common to all satellites. The terms left are ionospheric difference, multipath difference, and

random noises. See Figure 8.7. The pseudorange ionosphere-free function (bottom figure)

exhibits values around zero. The variations are multipath difference and random noise.

136
Figure 8. 7 OMC difference between satellites for P1 (top) and pseudorange Ionosphere-

free (bottom). The base satellite is PRN5 [WES2, DOY2(2000)]

8.3.8 Pseudorange Ionospheric Variations

Since the ionospheric effect on the GPS signals is frequency dependent, it is possible to

compute ionospheric variations by simply taking the difference of the pseudorange P1 and P2

from the same satellite at the same observation epoch (equation (5. 42)), detail given in section

5.3.4.2. Figure 8.8 shows ionospheric variation from pseudorange.

137
Figure 8. 8 Pseudorange ionospheric variations PRN5 [WES2, DOY2(2000)]

8.3.9 Elevation Angle and Azimuth

Part of the analysis is to have knowledge of where the observed satellite is located in the sky,

at what elevation and azimuth. The elevation and azimuth plots in Figure 8.9 show the

trajectory movement visually.

138
Figure 8. 9 Elevation and azimuth for PRN5 [WES2, DOY2(2000)]

8.3.10 Sky Plot

Sky plot provides a 3D sense of satellite trajectory centered at the receiver position. In the

Figure 8.10, the circular rings are 30 degrees apart.

139
Figure 8. 10 Sky plot for PRN5 (top) and all satellites (bottom) [WES2, DOY2(2000)]

140
8.3.11 North East & Up

The EKF yields receiver coordinate estimates which have been converted to positional

discrepancies in north, east, and up components with respect to the published coordinates.

Figure 8.11 is useful for judging convergence of the estimation.

Figure 8. 11 North (top), east (middle) and up (bottom) [WES2, DOY2(2000)]

141
8.3.12 Receiver Clock Error Estimate

WES2 station is equipped with a Hydrogen-Maser atomic clock. As a result, the receiver clock

estimate reveals a fairly smooth clock drift. Shown in Figure 8.12 are the receiver clock

estimate and its respective variance.

Figure 8. 12 Receiver clock estimate(top) and its variance (bottom) [WES2,

DOY2(2000)]

8.3.13 Ionosphere-free Pseudorange OMC (Innovation)

The EKF ionosphere-free pseudorange residuals in Figure 8.13 reflect multipath pattern seen

in Figure 8.3. All satellites exhibit similar residual magnitudes. These rather large values are

142
consequences of the linear combination which amplifies noise as well as multipath from P1

and P2.

Figure 8. 13 Ionosphere-free pseudorange OMC (innovation) for PRN5 (top) and all

satellites (bottom) [WES2, DOY2(2000)]

8.3.14 Ionosphere-free Carrier Phase OMC (Innovation)

In Figure 8.14, ionosphere-free carrier phase residuals illustrate shapes comparable to those

from pseudorange (Figure 8.13 ), but with much smaller magnitudes. Even though the linear

ionosphere-free carrier phase combination is able to eliminate ionosphere but it in turn

magnifies variance noises inherent in the L1 and L2 signals.

143
Figure 8. 14 Ionosphere-free carrier phase OMC (innovation) for PRN5 (top) and all

satellites (bottom) [WES2, DOY2(2000)]

8.3.15 Reparameterized Ambiguity Estimates

The ability to estimate real ambiguities is clearly shown in Figure 8.15. The ambiguities are

estimated every epoch and changed in the measurement update step due to a new observation.

After a few observation epochs the ambiguities seem to stabilize and converge. This

manifests the EKF algorithm performance.

144
Figure 8. 15 Reparameterized ambiguity estimates PRN5 (top) and all satellites (bottom)

[WES2, DOY2(2000)]

8.3.16 Variance of Estimated Ambiguity

The ambiguity variance stabilizes at a constant number. The log scale is used to aid

visualization in Figure 8.16.

145
Figure 8. 16 Variance of estimated ambiguity PRN5 (top) and all satellites (bottom)

[WES2, DOY2(2000)]

8.3.17 Number of Satellites Used in Kalman Filter and DOPs

The number of satellites available at each epoch and the satellite geometry as illustrated by the

DOPs provide valuable insight information for the analysis. This knowledge is particularly

important when problems happen due to singular satellite configuration (circular cone shape)

or lack of satellite clock. See Figure 8.17.

146
Figure 8. 17 Number of SV used in the computation (top) and the respective DOPs

(bottom) [WES2, DOY2(2000)]

8.3.18 Relativistic Correction

Relativistic effects on the observation depend on the position and speed of the satellite

(equation (2.22)), thus changes along the satellite trajectory. Relativistic correction for PRN5

is displayed in Figure 8.18. This correction does not cancel for PPP.

147
Figure 8. 18 Relativistic correction for PRN5 [WES2, DOY2(2000)]

8.3.19 Approximated Tropospheric Effect Using Saastamoinen Model

Environmental information at the observed station can be used to compute the approximate

tropospheric effect on the GPS signals using available tropospheric models. The

Saastamoinen tropospheric model has been chosen to calculate the zenith troposphere. Similar

to other models, the Saastamoinen tropospheric model requires information about temperature,

pressure, and relative humidity. The Niell Mapping Function (NMF) is then applied to

compute the effect at a specific satellite elevation. Figure 8.19 displays the plot of the

approximated tropospheric effect. The relation to elevation angle can readily be seen in

Figure 8.9.

Figure 8. 19 Approximated tropospheric error PRN5 (top) [WES2, DOY2(2000)]

148
8.3.20 Estimated Tropospheric Zenith Delay

The Zenith tropospheric delay is one of the parameters in the EKF. The NMF is applied and

used when computing the respective partial derivative in the H matrix. Figure 8.20 shows the

estimated zenith troposphere. The dip around epoch 200 corresponding to the peaks of the up

component (Figure 8.11) and receiver clock (Figure 8.12). This demonstrates the strong

correlation between troposphere, up, and receiver clock (see next section). This phenomenon

is due to the change in geometry as satellite PRN5 rises (see Figures 8.15 and 8.17). Zenith

tropospheric delay, up, and receiver clock are dependent upon satellite elevation angle and are

estimated well only when satellites are available on at least three distinctive elevations and

even then the precision of estimation is relatively weak (Vermeer, 1997).

Figure 8. 20 Estimated zenith tropospheric error [WES2, DOY2(2000)]

8.3.21 Correlation of the Estimated Parameters

Correlation coefficients of the estimated parameters are given in Table 8.4, for the last epoch

analyzed. The largest positive correlation is between the clock and up whereas the largest

negative correlation is between clock and troposphere.

149
Table 8. 4 Correlation coefficients of the estimated parameters [WES2, DOY2(2000)]

North East Up Clock Troposphere


North 1
East -0.09 1
Up -0.23 -0.35 1
Clock -0.01 0.28 0.64 1
Troposphere -0.13 -0.07 -0.30 -0.80 1

Figure 8.21 shows correlation coefficient for every epoch. The high correlation

between clock and up and clock and troposphere can readily be seen in the top figure. The

second figure from top shows the correlation between north, east, and up. The ambiguity

parameters appear to be highly correlated among themselves (second figure from bottom).

The bottom figure shows the correlation between ambiguity (PRN5) and other components

(troposphere, receiver clock, north, east, and up).

150
Figure 8. 21 Correlation coefficients between the estimated parameters: clock-up-

troposphere (top), north-east-up(second from top), ambiguities (second from

bottom), ambiguity-north-east-up-troposphere (bottom) [WES2, DOY2(2000)]

151
8.4 Experiments

Dual and single-frequency ionosphere-free linear combination (equations 5.40, 5.41 and 5.44)

analysis of all four stations using the PPP algorithm was made. The estimate coordinates

(after convergence) are compared with the published station coordinates. Solutions were

made with or without solid earth tide correction. The comparison was also made with

solutions obtained from JPL's automated GPS data analysis service (Zumberge, 1998). The

accuracy of JPL's PPP was verified by Witchayangkoon and Segantine (1999). Note that

JPL's PPP solutions correct for ocean loading tides by interpolating correction values from

known, nearby tide stations. The 3D ocean loading tides have magnitude of up to a few

centimeters. This study does not take ocean loading, receiver phase center offset, and phase

wind up angle into account.

8.4.1 Discrepancies With Respect to Published Coordinates

Overall the PPP algorithm developed for this research seems to perform better than JPL's

solutions after discontinuation of SA, 9.3 ± 3.0 cm versus 10.3 ± 2.2 (see Table 8.9). With

SA-on, JPL's solutions appear to perform better because JPL's software can handle the clock

better, 9.8 ± 2.2 cm versus 13.4 ± 2.7 cm. Many single-frequency solutions exhibit equivalent

accuracy as obtained from dual-frequency observations. However, single-frequency solutions

do not appear to be as robust as the dual-frequency. This is because pseudoranges which are

subjected to greater multipath dominate the single-frequency ionosphere-free combination.

Moreover, dual-frequency solutions seem to converge faster than single-frequency solutions,

particularly in the presence of high multipath and great widelane variations. This is due to the

additional geometric strength delivered from the second frequency. The differences between

the PPP solutions and the published coordinates are given in the Tables 8.5-8.9.

152
Table 8. 5 The difference between the PPP solutions and the published coordinates for
station WES2 (cm)

PPP
DOY PPP PPP (JPL)
+ Solid Earth Tides
(2000)
Dual L1 L2 Dual L1 L2 Dual

002 16.3 17.5 14.8 12 17.1 9.8 8.5

122 14.8 12.9 23.4 10.4 11.8 20.5 8.5

128 17.3 41.6 35.6 12.2 36 29.7 10.9

136 6.2 8.8 9.5 4.5 8.2 6.8 5.1

137 5.4 9.5 10.8 5.1 9.4 13.6 10.1

138 10.0 12.4 12.8 8.1 10.3 10.9 9.9

Table 8. 6 The difference between the PPP solutions and the published coordinates for
station NJIT (cm)

PPP
DOY PPP PPP (JPL)
+ Solid Earth Tides
(2000)
Dual L1 L2 Dual L1 L2 Dual
002 23.9 17.2 27.2 17 4.1 25.6 10.5

122 13 16.8 17.3 9.9 13.7 15.7 7.9

128 7.6 16.6 17.1 10.5 14.6 15.3 9.7

136 7.7 31 39.2 8.5 31.9 38.3 8.8

137 9.9 19.2 32.1 11.5 14.5 32.5 10.9

138 7.6 16.4 18.3 9.7 17.3 17.5 11.9

153
Table 8. 7 The difference between the PPP solutions and the published coordinates for
station USNO (cm)

PPP
DOY PPP PPP (JPL)
+ Solid Earth Tides
(2000)
Dual L1 L2 Dual L1 L2 Dual
100 8.6 10.7 10.5 14.2 13.2 15.8 7.8

122 12.2 18.7 20.2 12.4 16.5 18.1 8.9

128 15 23.4 14.5 12.9 20.1 16.5 11.3

137 2.7 17.9 16.6 6.1 18.8 19.6 8.8

138 11.7 19.6 11.9 11.1 19.3 16.1 11.5

Table 8. 8 The difference between the PPP solutions and the published coordinates for
station TSEA (cm)

PPP
DOY PPP PPP (JPL)
+ Solid Earth Tides
(2000)
Dual L1 L2 Dual L1 L2 Dual

065 11.2 8.9 9.5 14.9 16.1 15.9 12.2

122 19 18.2 9.7 16.8 22.2 12.9 13.7

136 13.5 15.3 13.6 13.8 16.7 16.3 14.4

138 10.1 14.1 20.3 7.4 14.5 15.9 11.2

Table 8. 9 Averaged discrepancies before and after SA-off (cm) for dual-frequency
solutions

SA Status PPP PPP (JPL)


SA-on 13.4 ± 2.7 9.8 ± 2.2
SA-off 9.3 ± 3.0 10.3 ± 2.2

154
8.4.2 Verification of Solid Earth Tides

Normally, solid earth tides cause site movement at the millimeter-level over a short period.

However, the absolute movement may be up to decimeters. From the experiment, it is found

that the PPP solution, more or less, reflects site displacements that correspond to the solid

earth tide corrections. An example of the WES2 station is shown in Figure 8.22. This

solution is from dual-frequency ionosphere-free observation, DOY2, 2000. Trends in north,

east, and up are indeed corresponding to each other, both the magnitude and the movement

directions.

Figure 8. 22 Station solution (top) compared with the solid earth tides corrections

(bottom) [WES2, DOY2(2000)]

155
8.4.3 Estimated Troposphere vs. Saastamoinen Model

The station WES2 normally provides a meteorology data file (m-file) accompanied with

observation and navigation files. Site environmental information includes station pressure

(mbar), temperature (degrees Celsius), relative humidity (percent), and sea level pressure

(mbar), recorded at 5-minute intervals. The pressure, temperature, and humidity sensors have

accuracy of 0.1 mbar, 0.5 degrees, and 5 percent, respectively. The station environmental data

is used to calculate zenith tropospheric delay using the Saastamoinen Model. The comparison

can be made between the estimated tropospheric zenith delay and the Saastamoinen model as

shown in Figure 8.23. The protuberance in the middle of troposphere estimated plot is caused

by satellite rising.

Figure 8. 23 Comparison of tropospheric zenith delay between the estimated value and

the Saastamoinen model [WES2, DOY136(2000)]

156
8.4.4 Impact of Satellite Antenna Offset

The satellite antenna offset cancels in differential GPS. PPP solution highly depends on

satellite antenna offset. Satellite antenna offset is given in the satellite coordinate system (see

Table 2.1 and section 2.4.7). Four cases are computed to verify the importance of satellite

antenna offset:

Case A: Take the satellite antenna offset into account following IGS convention,

 i  0.279
 j  =  0  m.
   
 k  1.023 

Case B: Ignore the satellite antenna offset,

 i  0
 j  = 0 m.
   
 k  0

Case C: Take only the satellite antenna i-offset,

 i  0.279
 j  =  0  m.
   
 k   0 

Case D: Take only the satellite antenna k-offset,

i  0 
 j  =  0  m.
   
 k  1.023

These four cases are tested using data from three stations from three different days. The

results (3D errors) are given in Table 8.10.

157
Table 8. 10 The impact of satellite antenna offset on the station solutions (3D errors)

DOY, Case (cm)


Station
2000 A B C D
WES2 002 16.3 24.9 17.3 19.8
NJIT 136 7.7 21.2 10.1 17.1
TSEA 138 10.1 17.6 10.3 17.1

Case C which takes only the satellite antenna i-offset into consideration gives slightly poorer

solutions than case A. The reason is that the satellite antenna k-offset points towards the

center of the Earth. The error committed by not considering the k-offset is mostly absorbed by

the receiver clock. As the satellite moves the i-offset, which points roughly towards the Sun,

is changing in the crust fix coordinate system. When the i-offset is not considered the solution

is significantly falsified and depends on the changing of direction of i.

8.4.5 Impact of Relativity

The relativity correction is important for PPP. If it is not applied, the PPP solution accuracy

will degrade to several meters. Table 8.11 shows the impact of ignoring relativistic effect.

Table 8. 11 Station coordinates offset due to ignoring relativistic effect

Station DOY, 2000 Offset (m)

WES2 002 7.25


WES2 137 11.60
TSEA 138 5.19
USNO 138 11.21

158
8.4.6 Impact of Single Cycle Slip

This section shows the case when there is a slip of one cycle embedded in the observation

data. This is simply done by putting the slip of one cycle into the carrier phase. A cycle slip

of one is added at epoch 300 to the original L1 PRN29 carrier phase data that span over six-

hour period [station USNO, DOY138]. The impact is seen in Figure 8.24. The 3D offset

increases to 49.8 cm. Also, one cycle slip is subtracted to the same data at the same epoch.

The impact displays in Figure 8.25. The 3D offset is 64.1 cm in this case.

For the WES2 [DOY2(2000)], a slip of one cycle is added at epoch 300 from original

L1 PRN4 data which spanned almost six hours. The accuracy reduces to 44.6 cm and 25 cm

for the positive and negative slip, respectively.

This simple experiment clearly demonstrates the importance of detecting cycle slips

for PPP.

159
Figure 8. 24 Adding a slip of one cycle to PRN29 at epoch 300 [USNO, DOY138(2000)]

160
Figure 8. 25 Subtracting a slip of one cycle to PRN29 at epoch 300 [USNO,

DOY138(2000)]

161
8.4.7 Using A Priori Tropospheric Information

Meteorological data (m-file) recorded at station WES2 provided an opportunity to experiment

with using this kind of information in PPP. It was expected that the solutions would converge

faster using a priori meteorological data. The meteorological data is used to compute zenith

tropospheric delay from the Saastamoinen model. This approximate troposphere is treated as

a priori information in the EKF analysis. The experiment is conducted on data sets from three

different days. The solutions are then compared with solutions obtained when estimating the

zenith troposphere parameter. In both cases, the NMF is used.

On DOY 137 and 138, it is apparent that the solutions using a priori tropospheric

information converge faster than when estimating the zenith tropospheric delay. However,

the solutions on DOY136 produce more or less the same convergence speed. These results are

inconclusive and require further study.

162
Figure 8. 26 Solutions comparison between zenith troposphere estimated and

approximated [WES2, DOY136(top), 137(middle), and 138(bottom) (2000)]

163
8.4.8 Kinematic Positioning

8.4.8.1 Dual-Frequency Kinematic Positioning

Data from two stations are used to investigate kinematic positioning using dual-frequency

pseudorange and carrier phase ionosphere-free functions. The zenith troposphere is either

estimated or computed from the Saastamoinen model using meteorological data. Even though

the data sets are from stationary receivers, the kinematic positioning is simulated by setting the

variances for the station coordinates to 1 m2 each epoch (and zero correlation). It is assumed

that the respective initial estimates of the coordinates will be available from a pre-analysis.

Both solutions yield more or less comparable results, the 3D discrepancy being about half a

meter and better most of the time. Utilizing meteorological information does not seem to

improve the solutions significantly (Figures 8.27 and 8.29). In all cases, horizontal accuracy

is about 0.2m, (Figures 8.28 and 8.30).

164
Figure 8. 27 Dual-frequency kinematic solutions comparison between troposphere

estimated and approximated (top) with the associated DOPs (middle) and SV

used (bottom) [WES2, DOY136(2000)]

165
Figure 8. 28 Horizontal offset comparison of kinematic solutions between zenith

troposphere estimated (top) and approximated (bottom) [WES2,

DOY136(2000)]

166
Figure 8. 29 Dual-frequency kinematic solutions comparison between troposphere

estimated and approximated (top) with the associated DOPs (middle) and SV

used (bottom) [USNO, DOY137(2000)]

167
Figure 8. 30 Horizontal offset comparison of kinematic solutions between zenith

troposphere estimated (top) and approximated (bottom) [USNO,

DOY137(2000)]

168
8.4.8.2 Single-Frequency Kinematic Positioning

In case of single-frequency kinematic positioning, the linear ionosphere-free combination is

half of the sum of the pseudorange and the carrier phase according to equation (5.44).

Solutions are made for the cases when the zenith troposphere is estimated and when it is

corrected from the meteorological data. The 3D discrepancies shown in Figure 8.31 indicate

that the solution obtained using a priori zenith troposphere information is the more precise and

possibly even a more accurate solution.

In case of horizontal offsets both solutions appear to be of the same quality (Figure

8.32). From this analysis, one might conclude that a priori tropospheric information will be

beneficial for the single-frequency user. The lack of geometric strength using just a single

frequency might be offset by the additional tropospheric data. More experimentation is

needed in order to confidently quantify single-frequency PPP in the kinematic mode.

169
Figure 8. 31 Single-frequency kinematic solutions comparison between troposphere

estimated (top) and approximated (second from top) with SV used (second from

bottom) and DOPs (bottom) [WES2, DOY2(2000)]

170
Figure 8. 32 Horizontal offset comparison of kinematic solutions between zenith

troposphere estimated (top) and approximated (bottom) [WES2, DOY2(2000)]

171
9 Conclusions and Recommendations

This thesis investigated GPS precise point positioning (PPP). Mathcad software was

developed to test various components of PPP and to serve as the software engine for further

studies. The PPP results were compared with published coordinates as well as with JPL's PPP

solutions available through the Internet. The single-frequency PPP has been developed,

tested, and compared with dual-frequency PPP solutions. Further, the correlation between the

up station coordinate, the zenith tropospheric delay, the receiver clock error, and the

ambiguities were studied. Moreover, the contribution of a priori tropospheric information was

investigated for static and kinematic positioning using either single or dual-frequency

observations.

9.1 Conclusions

The results of this research are documented by many figures and tables in Chapter 8. In all

cases the length of the data sets is 6 hours. The observation interval of 30 seconds is used

because the precise JPL ephemeris is currently only available at that rate. Several cases have

been identified that require further study.

172
9.1.1 Dual-Frequency Solutions

Excellent results have been obtained for the dual-frequency PPP solutions. First, consider the

solutions after SA has been discontinued. The research shows an average 3D discrepancy of

9.3 ± 3 cm while the JPL Internet service yields 10.3 ± 2.2 cm for the same data. See Table

8.9. This result might shed light on the significance (or insignificance) of the wind-up phase

correction. This correction was not applied in this research whereas JPL apparently uses such

a correction (Zumberge, private communication). In case of observations which are corrupted

by SA, the JPL solutions appear to be better (9.8 ± 2.2 cm versus 13.4 ± 2.7 cm ). The reason

for this reversal might be related to the interpolation of the satellite clock correction. In this

research the satellite clock correction was interpolated assuming that it changes continuously.

Zumberge (private communication) appears to indicate that they treat the satellite clock as a

step function, i.e. the satellite clock error should be considered constant during the duration of

the signal travel time (as a consequence of the manner it was originally computed by the

processing center). Because SA has been discontinued no attempts have been made to clarify

the clock correction situation. It is assume that the additional, effective satellite clock stability

gained by not having SA makes further investigation into satellite clock interpolation

unnecessary.

9.1.2 Impact of IGS Products and Service

Satellite position and satellite clock corrections are vital for PPP. Accuracy of the IGS

ephemeris and satellite clock correction has proven to be excellent, having a great impact on

PPP. When SA was active, the IGS products are very accurate. Even though the satellite

clock changed rapidly, SA did not reduce the accuracy of PPP.

173
There are occasionally epochs for which satellite clock corrections are not available.

The respective satellite observation cannot be used in PPP. This reduces the number of

satellites available, causes the DOPs to increase, and the accuracy of PPP to diminish. An

example is seen in Figures 8.11.

Normally the final IGS products are available about 2 weeks after the observation.

Clearly, if the data product could be made available faster, the PPP analysis could be executed

sooner.

9.1.3 Single-Frequency PPP

This study showed that in cases where an observation has low multipath variations single-

frequency ionosphere-free PPP solutions are equivalent to the dual-frequency solutions. This

conclusion provides great value to surveyors who mostly use single-frequency receivers.

However, like in the dual-frequency case, the observation must be free of cycle slips. Special

attention to cycle slip detection and fixing technique is required (see 9.2.1).

In the case of single-frequency solutions, pseudorange dominates the ionosphere-free

combination because of its large variances. This causes longer convergence times. In other

words, the single-frequency ionosphere-free function is essentially of P-code precision.

9.1.4 Correlation Between the Up Coordinate, Zenith Tropospheric Delay, Receiver

Clock, and Ambiguities

The experiments show a high correlation between receiver clock and the up coordinate. The

troposphere has a high negative correlation with receiver clock and up. However, the

troposphere is more correlated with the receiver clock than the up component (about three

174
times). Due to such high correlation among these three parameters undetected errors in the

satellite clock or cycle slips in the observations cause these three components to fluctuate

rapidly.

The ambiguities have little correlation with up, troposphere, and receiver clock, but

show high correlations among themselves. Estimated ambiguities appear to be steadier than

the other three components.

9.1.5 A Priori Tropospheric Information

Meteorological data is available at some observing stations. However, PPP analysis has

demonstrated the capability of reaching the same accuracy solutions at about 10 cm with or

without prior tropospheric input. Although the number of parameters being estimated is

reduced when tropospheric information is used, the expectation of faster convergence for dual-

frequency static PPP is inconclusive and needs further study.

9.1.6 Kinematic PPP

Dual-frequency kinematic PPP yields horizontal accuracy of about 20 cm, both with and

without priori tropospheric information. In addition, the 3D solutions give comparable results.

However, when meteorological information is used, the solutions tend to converge faster.

Single-frequency kinematic PPP seems to benefit from a priori tropospheric

information. The result seems to indicate that single-frequency users can expect PPP

positional accuracy at the 1 m level. Additional experimentation is required to determine the

suitability of inexpensive antennas often used in surveying or GIS applications.

175
9.2 Recommendations

9.2.1 Cycle Slip

PPP requires cycle slip free data. A good and robust algorithm to detect and fix cycle slip in

undifferenced observations is necessary prior to running the Kalman filter. In this research the

wide lane and the ionosphere-free carrier phase function were used to detect and remove slips

using a semi-graphical interactive procedure. In view of the discontinuation of SA it might be

advantageous to analyze the between-satellite OMC for cycle slip detection and removal. The

latter approach is particularly important for single-frequency users. In fact, the

discontinuation of SA might be most beneficial to the single-frequency PPP user.

Experimentation should also be done analyzing the estimated ambiguities for missed cycle

slips and re-running the solution with corrected observations.

9.2.2 Phase Wind-up Error

For a given satellite the phase wind-up error may be as much as a cycle and changing

systematically with the satellite path. Its effect on the position is possibly reduced when

averaging over satellites (Hugentobler, private communication). Correcting the phase wind-

up error is expected to improve mostly the receiver clock estimates. Incorporation of the

phase wind-up correction into PPP needs to be studied further. It seems that JPL has

implemented a wind-up algorithm and that the Bernese group is in the stage of

experimentation with the wind-up correction.

176
9.2.3 Receiver Antenna Phase Center Offset

Receiver phase center offset and variation has not been taken into account in this research.

According to Mader (1999) the receiver antenna phase center offset can cause errors in the up

coordinate by as much as 10 cm and sub-centimeter errors in the horizontal. The phase center

variation is a function of both elevation and azimuth. Phase center offset and variation, of

course, are antenna specific. It is interesting to note that the 3D comparisons with respect to

the published antenna coordinates showed the largest discrepancy in height. Because the

height component is sensitive to several error sources this observation is probably not

conclusive. Also, any error in the vertical L1 & L2 phase center offset might be absorbed by

the receiver clock estimate. According to the CORS Internet document the published

coordinates of the stations refer to the L1 phase center.

9.2.4 A Priori Tropospheric Information

This research found that a priori tropospheric information is important for single-frequency

kinematic PPP because it appears to add strength to the solution (by not having to estimate the

troposphere). Such an observation could not be confirmed with dual-frequency observations.

The use of a priori tropospheric information to reduce the time of convergence requires

additional study.

9.2.5 GLONASS

Several IGS processing centers are now producing the combined GPS/GLONASS precise

ephemeris in SP3 format. This study should be extended to include GLONASS observations

to explore their contribution to PPP. Since the PPP technique is not concerned with fixing

177
ambiguities the additional difficulties caused by the GLONASS frequencies are not relevant.

However, an additional receiver clock bias parameter will be needed.

178
REFERENCES

Aloi, D. N. 1999. Phase Center Variation (PCV) Determination of the Ohio University

Dipole Array Using GPS Data. Proc. ION GPS-99, Nashville, 705-712.

Altamimi, Z. 2000. Private Communication. Institut Geographique National,

ENSG/LAREG, France.

Argus, D. F. and M. B. Helflin. 1995. Plate Motion and Crustal Deformation Estimated with

Geodetic Data from the Global Positioning System. Geophys. Res. Lett., 22(15),

1973-1976, August.

Bar-Sever, Y. E. 1996. A New Model For GPS Yaw Attitude. J. Geodesy, 70, 714-723,

Springer-Verlag.

Bassiri, S. and G. A. Hajj. 1992. Modeling the Global Positioning System Signal Propagation

Through the Ionosphere. Telecommunications and Data Acquisition (TDA) Progress

Report 42-110, 92-103. August.

Bean, B. R. and E. J. Dutton. 1966. Radio Meteorology. National Bureau of Standards

Monograph 92. US Department of Commerce, 435p.

Beutler, G., I. I. Mueller, and R. E. Neilan 1994. The International GPS Service for

Geodynamics (IGS): Development and Start of Official Service on January 1, 1994.

Bull. Geod., 68(1), 43-46.

Blewitt, G. 1990. An Automated Editing Algorithm for GPS Data. Geophys. Res. Lett.,

17(3), 199-202, March.

179
Boucher, C. and Z. Altamimi. 1996. ITRF and Its Relationship to GPS. GPS World, 7(9),

September. Or http://lareg.ensg.ign.fr/ITRF/ITRF-GPS.html. Accessed May, 1999.

CDDIS. 2000a. Crustal Dynamics Data Information System (CDDIS) Server.

ftp://cddisa.gsfc.nasa.go/pub/igex. Accessed July 21, 2000.

CDDIS. 2000b. Crustal Dynamics Data Information System (CDDIS) Server.

ftp://cddisa.gsfc.nasa.gov/pub/gps/gpsdata. Accessed July 21, 2000.

Chao, C. C. 1974. The Tropospheric Calibration Model for Mariner Mars 1971. JPL

Technical Report 32-1587, 61-76.

Cheney, W. and D. Kincaid. 1994. Numerical Mathematics and Computing. 3rd ed.,

Brook/Cole Publishing Company(ITP), 578p.

Collins, J.P. and R.B. Langley. 1999. Possible Weighting Schemes for GPS Carrier Phase

Observations in the Presence of Multipath. Contract report for the United States Army

Corps of Engineers Topographic Engineering Center, No. DAAH04-96-C-0086 / TCN

98151, March.

Davis, J. L., T. A. Herring, I. I. Shapiro, A. E. E. Rogers, and G. Elgered. 1985. Geodesy by

Radio Interferometry: Effects of Atmospheric Modeling Errors on Estimates of

Baseline Length. Radio Sci., 20(6), November-December, 1593-1607.

DeMets, C., R. G. Gordon, D. F. Argus, and S. Stein. 1994. Effect of Recent Revisions to the

Geomagnetic Reversal Time Scale on Estimates of Current Plate Motions. Geophys.

Res. Lett., 21, 2191-2194.

Euler, H.-J. and C.C. Goad. 1991. On Optimal Filtering of GPS Dual Frequency Observations

Without Using Orbit Information. Bulletin Geodesique, 65(2), 130-143.

Gelb, A. (Ed) 1974. Applied Optimal Estimation. MIT Press, 14th printing (1996), 374p.

180
Goad, C. C. 1993. IGS Orbit Comparisons. Proc. the 1993 IGS Workshop, 218-225, Univ.

of Bern.

Gurtner, W. 1997. RINEX: The receiver Independent Exchange Format Version 2.

Astronomical Institue Univ. of Berne.

ftp://igscb.jpl.nasa.gov/igscb/data/format/rinex2.txt.

Hartinger, H. and F. K. Brunner. 1999. Variance of GPS Phase Observations: the Sigma-ε

Model. GPS Solutions, 2(4), Spring, 35-43.

Hein, G. W. 2000. From GPS and GLONASS via EGNOS to Galileo-Positioning and

Navigation in the Third Millennium. GPS Solutions, 3(4), Spring, 39-47.

Henriken, J., G. Lachapelle, J. Raquest, and J. Stephen. 1996. Analysis of Stand-Alone GPS

Positioning Using Post-Mission Information. Proc. ION GPS-97, Kansas City,

Missouri, 251-259.

Herring, T. A. 1992. Modeling Atmospheric Delays in the Analysis of Space Geodetic Data.

Proc. the Symposium: Refraction of the Transatmospheric Signals in Geodesy, the

Hague, the Netherlands.

Hé roux, P. and J. Kouba. 1995. GPS Precise Point Positioning with a Difference.

Geomatics'95, Ottawa, Ontario, Canada, June 13-15.

HNHA. 2000. Location of Plate Boundaries.

http://volcano.und.nodak.edu/vwdocs/vwlessons/plate_tectonics/part12.html.

Accessed July 21, 2000.

Hofmann-Wellenhof, B., H. Lichtenegger, and J. Collins. 1997. GPS: Theory and Practice.

4th, revised ed., SpringerWienNewYork, 389p.

Hopfield, H.S. 1969. Two-Quartic Tropospheric Refractivity Profile for Correcting Satellite

Data. J. Geophys. Res., 74(18), 4487-4499.

181
Hugentobler, U. 2000. Private Communication. Astronomical Institute, University of Berne,

Switzerland.

ICD-GLONASS. 1998. GLONASS Interface Control Document. Version 4.0. Coordination

Scientific Information Center, the Ministry of Defense of the Russian Federation, 45p.

ICD-GPS-200C. 1997. Navstar GPS Space Segment / Navigation User Interfaces: Interface

Control Document ICD-GPS-200C. Revision: IRN-200C-002. ARINC Research

Cooperation, CA.

IERS Conventions. 1996. IERS Technical Note 21. McCarthy, D. D. (Ed.) Observatoire de

Paris, July, 95p.

IERS96. 2000. The 1996 IERS Conventions Homepage: the Chapters, Subroutines, and

Associated Information. http://maia.usno.navy.mil/conventions.html. Accessed July

21, 2000.

IGEB. 2000. GPS Fluctuations Over Time on May 2, 2000.

http://www.igeb.gov/sa/diagram.shtml. Accessed July 21, 2000.

IGLOS-PP. 2000. International GLONASS Service Pilot Project (IGLOS-PP) Official

Homepage. http://igscb.jpl.nasa.gov/projects/iglos/glonass.html. Accessed July 21,

2000.

IGS. 1997. IGS Resource Information. September. Published by IGS Central Bureau.

IGS. 1998. IGS Directory. January. Published by IGS Central Bureau.

IGS. 2000. About the IGS. http://igscb.jpl.nasa.gov. or

http://igscb.jpl.nasa.gov/overview/viewindex.html. Accessed July 2000.

Janes, H. W. and R. B. Langley. 1989. A Comparison of Several Models for the Prediction of

Tropospheric Propagation Delay. Proc 5th Int. Symp. on Satellite Positioning, Las

Cruces, New Mexico, 777-788.

182
Janes, H. W., R. B. Langley, and S. P. Newby. 1991. Analysis of Tropospheric Delay

Prediction Models: Comparison with Ray-Tracing and Implication for GPS Relative

Positioning. Bull. Geod., 65, 151-161.

Jonkman, N. F. and K. de Jong. 2000a. Integrity Monitoring of IGEX-98 Data, Part I:

Availability. GPS Solutions, 3(4), Spring, 10-23.

Jonkman, N. F. and K. de Jong. 2000b. Integrity Monitoring of IGEX-98 Data, Part II: Cycle

Slip and Outlier Detection. GPS Solutions, 3(4), Spring, 24-34.

Jonkman, N. F. and K. de Jong. 2000c. Integrity Monitoring of IGEX-98 Data, Part III:

Broadcast Navigation Message Validation. GPS Solutions, 3(5), Summer.

JPL. 2000a. JPL Navigation Server. ftp://navigator.jpl.nasa.gov/pub/ephem/export.

Accessed July 21, 2000.

JPL. 2000b. GPS Time Series. http://sideshow.jpl.nasa.gov/mbh/series.html. Accessed July

21, 2000.

JPL. 2000c. JPL's High-Rate Ephemerides Database.

ftp://sideshow.jpl.nasa.gov/pub/jpligsac/hirate. Accessed January-July, 2000.

Jursa, A. S. (ed.) 1985. Handbook of Geophysics and the Space Environment. Air Force

Geophysics Laboratory, National Technical Information Services, Springfield,

Virginia.

Klobuchar, J. A. and P. H. Doherty. 1998. A Look Ahead: Expected Ionospheric Effects on

GPS in 2000. GPS Solutions, 2(1), 42-48.

Knight, M. and A. Finn. 1996. The Impact of Ionospheric Scintillations on GPS

Performance. Proc. ION-GPS 96, Kansas City, 555-564.

Kouba, J., Y. Mireault, G. Beutler, T. Springer, and G. Gendt. 1998. A Discussion of IGS

Solutions and Their Impact on Geodetic and Geophysical Applications. GPS

Solutions, 2(2), 3-15.

183
Kouba, J., and T. Springer. 1998. Satellite Antenna Offsets, etc. E-mail communications

Among IGS Analysis Centers. http://maia.usno.navy.mil/gpst/mail/ant-offs Accessed

Jan-August, 2000.

Kouba, J. 2000. IGS Orbit/Clock Navigation & Satellite Clock Interpolation With No SA.

IGSMAIL-2824. http://igscb.jpl.nasa.gov/mail/igsmail/2000/msg00163.html.

Accessed August, 2000.

Kovalevsky, J., and I. I. Mueller. 1989. Introduction. Reference Frames in Astronomy and

Geophysics, J. Kovalevsky, I. I. Mueller, and B. Kolaczek (Eds.), Kluwer Academic

Publishers, Netherlands.

Kunches, J. M. and J. A. Klobuchar. 1998. Eye on the Ionosphere. GPS Solutions, 2(1), 66-

66.

Lachapelle, G., M. E. Cannon, W. Qiu, and C. Varner. 1996. Precise Aircraft Single-Point

Positioning Using GPS Post-Mission Orbits and Satellite Clock Corrections. J.

Geodesy, 70, 562-571, Springer-Verlag.

Lambeck, K. 1980. The Earth’s Variable Rotation: Geophysical Causes and Consequences.

Cambridge Univ. Press, England, 449p.

Lambeck, K. 1989. The Fourth Dimension in Geodesy Observing the Deformation of the

Earth. Lecture Notes in Earth Science (29): Developments in Four-Dimensional

Geodesy, F. K. Brunner and C. Rizos (Eds.), Spring-Verlag, 1-14.

Langley, R. B. 1996. GPS Receivers and the Observables. Lecture Notes in Earth Sciences

(60): GPS for Geodesy, A. Kleusberg and P. J. G. Teunissen (Eds.), Spring-Verlag,

141-173.

Langley, R. B. 1997. GPS Receiver System Noise. GPS World, 8(6), 40-45.

Langley, R. B. 1999. Dilution of Precision. GPS World, 10(5), 52-59.

184
Lanyi, G. 1984. Tropospheric Delay Effects in Radio Interferometry. The

Telecommunications and Data Acquisition (TDA) Progress Report 42-78, Imprint:

JPL, California Inst. Tech., 152-159.

Laurila, S. H. 1976. Electronic Surveying and Navigation. Wiley-InterScience, 545p.

LAREG. 2000a. ITRF. http://lareg.ensg.ign.fr/ITRF. Accessed June, 2000.

LAREG, 2000b. ITRF96. http://lareg.ensg.ign.fr/ITRF/itrf96.html Accessed June, 2000.

LAREG, 2000c. ITRF97. http://lareg.ensg.ign.fr/ITRF/ITRF97 Accessed June, 2000.

LAREG, 2000d. ITRF2000: A Call for Participation. http://lareg.ensg.ign.fr/ITRF/ITRF2000

Accessed June, 2000.

Leick, A. 1995. Satellite GPS Surveying. 2nd ed., Wiley-InterScience, 560p.

Leick, A. and C. Mundo. 1997. GLONASS Double Difference Ambiguities. Intl Symposium

on Current Crustal Movement and Hazard Reduction, Wuhan, P.R.China, Nov 4-7.

Leick, A., J. Beser, and P. Rosenboom. 1998a. Aspects of GLONASS Carrier-Phase

Differencing. GPS Solutions, 2(1), Summer, 36-41.

Leick, A., J. Beser, P. Rosenboom, and B. Wiley. 1998b. Assessing GLONASS Observation.

Proc. ION GPS-98, Nashville, Tennessee.

Li, J. 1995. OTF With GLONASS Carrier Phases. M.S. Thesis. Univ. Maine. 128p.

Mader, G. L. 1999. GPS Antenna Calibration at the National Geodetic Survey. GPS

Solutions, 3(1), 50-58.

Malys, S. and J. A. Slater. 1994. Maintenance and Enhancement of the World Geodetic

System 1984. Proc. ION GPS-94, Salt Lake City, Utah.

Marini, J.W. 1972. Correction of Satellite Tracking Data for an Arbitrary Tropospheric

Profile. Radio Sci., 7(2), 223-231.

185
Marini, J.W. and C.W. Murray. 1973. Correction of Satellite Tracking Data for Atmospheric

Refraction at Elevations above 10 degrees. NASA Report X-591-73-351, Goddard

Space Flight Center.

Melchior, P. 1978. The Tides of the Planet Earth. Pergamon Press, p.609.

Melchior, P. 1983. The Tides of the Planet Earth. 2nd ed., Pergamon Press, p.641.

Mendes, V. B. and R. B. Langley. 1994. A Comprehensive Analysis of Mapping Functions

Used in Modeling Tropospheric Propagation Delay in Space Geodetic Data. KIS94,

Proc. of the Int. Symp. on Kinematic Systems in Geodesy, Geomatics and Navigation,

Banff, Alberta, 30 August - 2 September 1994, The University of Calgary, Calgary,

Alberta, 87-98.

Mendes, V. B. and R. B. Langley. 1998. Optimization of Tropospheric Delay Mapping

Function Performance for High-Precision Geodetic Applications. Proc. DORIS Days,

April 27-29, Toulouse, France.

Mendes, V. B. 1998. Modeling the neutral-atmosphere propagation delay in radiometric

space techniques. Ph.D. Thesis. Univ. New Brunswick, p.353.

Mertikas, S. P. and C. Rizos. 1997. On-line Detection of Abrupt Changes in the Carrier-

Phase Measurement of GPS. J. Geodesy, 71, 469-482, Springer-Verlag.

Misra, P. N., R. Abbot, and E. Gaposchkin. 1996. “Integrated Use of GPS and GLONASS:

Transformation between WGS 84 and PZ-90.” Proc. ION GPS-96, Kansas City.

Musman, S., G. Mader, and C. E. Dutton. 1998. Total Electron Content Changes in the

Ionosphere During the January 10, 1997 Disturbance. Geophys. Res. Lett., 25(15),

August, 3055-3058.

NASA. 2000a. U.S. Standard Atmosphere 1976.

http://nssdc.gsfc.nasa.gov/space/model/atmos/us_standard.html. Accessed June,

2000.

186
NASA. 2000b Sunspot cycle prediction.

http://science.msfc.nasa.gov/ssl/pad/solar/predict.htm. Accessed June 30, 2000.

Navcen. 2000. U.S. Coast Guard Navigation Center. http://www.navcen.uscg.mil Accessed

July 21, 2000.

Neilan, R. E., J. F. Zumberge, G. Beutler, and J. Kouba. 1997. The International GPS

Service: A Global Resource for GPS Applications and Research. Proc. ION GPS-97,

Kansas City, Missouri, September.

NGS. 2000. The NGS GPS Orbital Formats. http://cddisa.gsfc.nasa.gov/dgps/orbits.txt

Niell, A. E. 1996. Global Mapping Functions for the Atmosphere Delay at Radio

Wavelengths. J. Geophys. Res., 101(B2), 3227-3246.

NIMA. 1997. Department of Defense World Geodetic System 1984: Its Definition and

Relationships with Local Geodetic Systems. Technical Report, NIMA TR8350.2, 3rd

ed., July.

NIMA. 2000. SP3 Format Description. http://164.214.2.59/GandG/sathtml/sp3format.html.

Accessed August, 2000.

NOAA. 2000. Sunspot Numbers. http://www.ngdc.noaa.gov/stp/SOLAR/SSN/ssn.html.

Accessed July 22, 2000.

NOAA/NASA/USAF. 1976. U.S. Standard Atmosphere, 1976. U.S. Government Printing,

Washington DC.

Nordwall, B. 1996. Atmospheric/Multipath Concerns for DGPS. Aviation Week & Space

Technology, October, 60.

OSO, 2000. Ocean Loading Database. Onsala Space Observatory, Chalmers University of

Technology. ftp://gere.oso.chalmers.se/pub/hgs/oload/olrs18.blq. Accessed June 15,

2000.

187
Ray, J. R. 1998. The IGS/BIPM Time Transfer Project. Proc. the IGS 1998 Analysis Center

Workshop, Darmstadt, Germany, February 9-11.

Remondi, B.W. 1993. NGS Second Generation ASCII and Binary Orbit Formats and

Associated Interpolation Studies. IAG Symposia (Symp. 109, Vienna, Austria, August

11-14, 1989) by IUGG, and IAG, G. L. Mader (Ed.). Springer-Verlag.

Saasamoinen, J. 1971. Atmospheric Correction for the Troposphere and Stratosphere in

Radio Ranging of Satellite. Int. Symp. on the Use of Artificial Satellite, Henriksen

(ed.), 3rd Washington, 247-251.

Scherneck H. G. and F. H. Webb. 1998a. Ocean Tide Loading and Diurnal Tidal Motion of

the Solid Earth Center. IERS (1998) Technical Note No. 25.

Scherneck H. G., R. Haas, and F. H. Webb. 1998b. Ocean Loading Tides in Space

Techniques and Implications for the Mass Center Variations. Proc. 13th Gen. Meeting

Nordic Geodetic Comm. Gvle 1998, B. Jonsson (Ed.).

SFCSIC. 2000. Coordination Scientific Information Center (KNITs) of the Ministry of

Defense of the Russian Federation. http://mx.iki.rssi.ru/SFCSIC/SFCSIC_main.html.

Accessed July 21, 2000.

Slater, J. A. 2000. IGLOS Electronic Mail. Message Number 1, May, 25. Available at

http://igscb.jpl.nasa.gov/mail/iglosmail/2000/maillist.html. Accessed July 21, 2000.

Slater, J. A., C. E. Noll, and K. T. Gowey (Editors). 2000. International GLONASS

Experiment IGEX-98. Workshop Proc. September, 13-14, 1999. Published jointly by

IGS Central Bureau and Jet Propulsion Laboratory (California Institute of

Technology), May 2000.

Springer, T. A., J. F. Zumberge, and J. Kouba. 1998. The IGS Analysis Products and the

Consistency of the Combined Solutions. 1998 Analysis Center Workshop Proc.,

Darmstadt, Germany, February 9-11, 37-54.

188
Tetewsky, A. K. and F. E. Mullen. 1997. Carrier Phase Wrap-Up Induced by Rotating GPS

Antennas. GPS World, Innovation column, 51-57, February.

Teunissen, P. J. G. 1994. A New Method for Fast Carrier Phase Ambiguity Estimation. Proc.

IEEE PLANS94, 562-574.

Teunissen, T. J. G. 1997. On the GPS Widelane and Its Decorrelating Property. J Geodesy,

71(9), 577-587.

Teunissen, T. J. G. 1998. Weighting GPS Dual Frequency Observations: Bearing the Cross of

Cross-Correlation. GPS Solutions, 2(2), 28-37.

Tralli, D. M. and S. M. Lichen. 1990. Stochastic Estimation of Tropospheric Path Delays in

Global Positioning System Geodetic Measurements. Bull. Geod., 64, 127-159.

Tregoning P., B. Twilley, M. Hendy, and D. Zwartz. 1998. Monitoring Isostatic Rebound in

Antarctica Using Continuous Remote GPS Observations. GPS Solutions, 2(3),

Winter.

USNO. 2000a. GPS Time Transfer. http://tycho.usno.navy.mil/gpstt.html. Accessed May,

2000.

USNO. 2000b. NAVSTAR GPS Operations. http://tycho.usno.navy.mil/gpsinfo.html.

Accessed May, 2000.

Vaní¹ek, P. and E. Krakiwsky. 1982. Geodesy: The Concepts. North-Holland Publishing

Company, p.691.

Vermeer, M. 1997. The Precision of Geodetic GPS and One Way of Improving It. J.

Geodesy, 71, 240-245, Springer-Verlag.

Wang, J., C. Rizos, M. P. Stewart, and A. Leick. 2000. GPS and GLONASS Integration:

Modeling and Ambiguity Resolution Issues. GPS Solutions (accepted for

publication).

189
Welch, G. and G. Bishop. 1997. An Introduction to Kalman Filter. Department of Computer

Science, Univ. North Carolina at Chapel Hill. TR-95-041.

http://www.cs.unc.edu/~welch/publications.html. Accessed May 1999.

Wilson, B. D. 1999. New broadcast Tgd values as April 29, 1999. IGS Electronic Mail No

23042304, June 11. http://maia.usno.navy.mil/gpst/mail/11Jun99.1. Accessed April,

2000.

Witchayangkoon, B. and P.C.L. Segantine. 1999. Testing JPL's PPP Service. GPS Solutions,

3(1), 73-76.

Wu, J. T., S. C. Wu, G. A. Hajj, W. I. Bertiger, and S. M. Lichen. 1993. Effects of Antenna

Orientation on GPS Carrier Phase. Manuscripta Geodaetica, 18, 91-98.

Wubbena, G., M Schmitz, F. Menge, G. Seeber. 1997. A New Approach for Field

Calibration of Absolute GPS Antenna Phase Center Variations. Navigation: J. the

Inst. of Navigation, 44(2), Summer, 247-255.

Zlotnicki, V. 1996. Which Tide? (Some definitions of Pure Ocean Tide, Earth (Body and

Loading) Tide, Long Period Tide, Internal Tide, Atmospheric Tide and Pole Tide).

http://podaac.jpl.nasa.gov/cdrom/tide/Document/text/tide_is.txt

Zumberge, J. F., M. B. Heflin, D. C. Jefferson, M. M. Watkins, and F. H. Webb. 1997a.

Precise Point Processing for the Efficient and Robust Analysis of GPS Data from

Large Networks. J. Geophys. Res., 102(B3), 5005-5017.

Zumberge, J.F., M. M. Watkins, and F. H. Webb. 1997b. Characteristics and Application of

Precise GPS Clock Solutions Every 30 Seconds. Navigation: J. the Inst. of

Navigation, 44(4), 449-456, Winter 1997-1998.

Zumberge, J. F. 1998. Automated GPS Data Analysis Service. GPS Solutions, 2(3), 76-78.

190
Appendix A: SP3 EPHEMERIS FORMAT

The SP3 format description given below is obtained from NIMA (2000). It is documented
here because the SP3 ephemeris (and its format) is central to PPP.

The SP3 format is presented here in both the position (P) mode and the velocity (V)
mode. The format was finalized by the National Geodetic Survey. The SP3 format is precise
to 1mm and 1 picosecond. If velocity is included, its precision is 10E-4 mm/sec and 10E-4
picoseconds/s.

I. Version Identification:
On line one, character two, there is now a single version identification character. The first
released version has been designated version 'a'. Subsequent versions will use the lower case
letters in alphabetical order.

II. Position/Velocity Mode Flag:


The proposed formats did not accommodate velocity data on the theory that it can be derived
from the positional data. Although it is true that velocity data can be computed from
positional data--accurate to a few microns per second or better--there are arguments why
velocity will be needed, explicitly, in some situations. On line one, character three, there is a
single flag to indicate if the SP3 file will comprise only positional data (P) or both position
and velocity data (V).

III. Position/Velocity Mode P:


The first character of each line is 'P' for position.

IV. Position/Velocity Mode V:


When the position/velocity mode flag is set to 'V', each position line for a given satellite is
followed by a velocity line for the same satellite. The velocity components are given in
decimeters/s and have a precision of 10E-14 mm/sec. The last column of a velocity line is the
rate of change of clock correction given in units of 10E-4 microsec/s. The precision of this
parameter is 10E-16 s/s.

A marker is designates the locations of clock events. This marker is an 'E' located in column
75 of the position line for a satellite. This field will normally be blank.

The following is the format for the SP3:

191
SP3 Format:

Line One:
('#aV',i4,4i3,f12.8,6x,'96 DD+AD WGS84 FIT NIMA')
Column1 Symbol #
2 Version Identifier a
3 Pos/Vel Mode Flag P or V
4-7 Year Start 1995
8 Unused _
9-10 Month Start _7
11 Unused _
12-13 Day of Month Start _6
14 Unused _
15-16 Hour Start _0

192
17 Unused _
18-19 Minute Start _0
20 Unused _
21-31 Second Start _0.00000000
32 Unused _
33-39 Number of Epochs _____96
40 Unused _
41-45 Data Used DD+AD
46 Unused _
47-51 Coordinate System WGS84
52 Unused _
53-55 Orbit Type FIT
56 Unused _
57-60 Agency NIMA

Line Two:
('##',i5,f16.8,f15.8,i6,f16.13)
Column1-2 Symbols ##
3 Unused _
4-7 GPS Week _808
8 Unused _
9-23 Seconds of Week 345600.00000000
24 Unused _
25-38 Epoch Interval __900.00000000
39 Unused _
40-44 Mod Julian Day Start 49904
45 Unused _
46-60 Fractional Day 0.0000000000000

Line Three:
('+',i5,3x,17i3)
Column1-2 Symbols _+
3-4 Unused __
5-6 Number of Sats 25
7-9 Unused ___
10-12 Sat #1 Id __1
13-15 Sat #2 Id __2
* * *
* * *
* * *
58-60 Sat #17 Id _22

Line Four:
('+',8x,17i3)
Column1-2 Symbols _+
3-9 Unused _______
10-12 Sat #18 Id _23
13-15 Sat #19 Id _24
* * *
* * *
* * *
58-60 Sat #34 Id __0

Line Five:
('+',8x,17i3)
(Same as Line Four for Sats 35 to 51)

193
Line Six:
('+',8x,17i3)
(Same as Line Four for Sats 52 to 68)

Line Seven:
('+',8x,17i3)
(Same as Line Four for Sats 69 to 85)

Line Eight:
('++',7x,17i3)
Column1-2 Symbols ++
3-9 Unused _______
10-12 Sat #1 Accuracy __0
13-15 Sat #2 Accuracy __0
* * *
* * *
* * *
58-60 Sat #17 Accuracy __0

Line Nine:
('++',7x,17i3)
(Same as Line Eight for Sats 18 to 34)

Line Ten:
('++',7x,17i3)
(Same as Line Eight for Sats 35 to 51)

Line Eleven:
('++',7x,17i3)
(Same as Line Eight for Sats 52 to 68)

Line Twelve:
('++',7x,17i3)
(Same as Line Eight for Sats 69 to 85)

Lines Thirteen and Fourteen:


('%c cc cc ccc ccc cccc cccc cccc cccc ccccc ccccc ccccc ccccc')
Column1-2 Symbols %c
3 Unused _
4-5 2 characters cc
6 Unused _
7-8 2 characters cc
9 Unused _
10-12 3 characters ccc
13 Unused _
14-16 3 characters ccc
17 Unused _
18-21 4 characters cccc
22 Unused _
23-26 4 characters cccc
27 Unused _
28-31 4 characters cccc
32 Unused _
33-36 4 characters cccc
37 Unused _

194
38-42 5 characters ccccc
43 Unused _
44-48 5 characters ccccc
49 Unused _
50-54 5 characters ccccc
55 Unused _
56-60 5 characters ccccc

Lines Fifteen and Sixteen:


('%f 0.0000000 0.000000000 0.00000000000 0.000000000000000')
Column1-2 Symbols %f
3 Unused _
4-13 10 column float _0.0000000
14 Unused _
15-26 12 column float _0.000000000
27 Unused _
28-41 14 column float _0.00000000000
42 Unused _
43-60 18 column float _0.000000000000000

Lines Seventeen and Eighteen:


('%i 0 0 0 0 0 0 0 0 0')
Column1-2 Symbols %i
3 Unused _
4-7 4 column integer ___0
8 Unused _
9-12 4 column integer ___0
13 Unused _
14-17 4 column integer ___0
18 Unused _
19-22 4 column integer ___0
23 Unused _
24-29 6 column integer _____0
30 Unused _
31-36 6 column integer _____0
37 Unused _
38-43 6 column integer _____0
44 Unused _
45-50 6 column integer _____0
51 Unused _
52-60 9 column integer ________0

Lines Nineteen to Twenty-two:


('/* CCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCC')
Column1-2 Symbols /*
3 Unused _
4-60 Comments CC...CC

Line Twenty-three (The Epoch Header Record):


('*',i6,4i3,f12.8)
Column1-2 Symbol *_
3 Unused _
4-7 Year Start 1995
8 Unused _
9-10 Month Start _7
11 Unused _

195
12-13 Day of Month Start _6
14 Unused _
15-16 Hour Start _0
17 Unused _
18-19 Minute Start _0
20 Unused _
21-31 Second Start _0.00000000

Line Twenty-four (The Position and Clock Record):


('P',i3,4f14.6)
Column1 Position P
2-4 Vehicle Id __3
5-18 x coordinate (km) ___5675.974300
19-32 y coordinate (km) _-20087.366600
33-46 z coordinate (km) __15925.186400
47-60 clock (microseconds) ____390.882890

Line Twenty-five (The Velocity Record):


('V',i3,4f14.6)
Column1 Velocity V
2-4 Vehicle Id __3
5-18 x coordinate (dm/sec) __11770.882098
19-32 y coordinate (dm/sec) _-17082.412288
33-46 z coordinate (dm/sec) _-25886.520572
47-60 Rate of change of _____-0.018200
Clock Correction (0.0001 microsec/sec)

(Line Twenty-four will include an 'E' in column 75 to designate an


Event in the Position and Clock Record.) ('P',i3,4f14.6,14x,'E')

(Lines Twenty-four and Twenty-five repeat for each vehicle in the


epoch.)

(Epoch Headers in Line Twenty-three repeat for each new epoch with
vehicle position and velocity records following.)

(Last line in file) EOF

Note that not available data is set 999999.999999.

196
Appendix B: RELEVANT MATHCAD FUNCTIONS

B.1 Cycle Slip Detection and Fixing

197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
B.2 OMC Computations

217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
B.3 Graphing and Data Editing

238
239
240
241
242
243
244
B.4 Kalman Estimation

245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
Appendix C: PSEUDORANGE PERFORMANCE USING PRECISE

EPHEMERIS

The pseudorange ionosphere-free solutions using precise ephemeris were carried out for two
days, applying meteorological data (Saastamoinen model) or estimating the troposphere.
Figure C.1 shows that the solutions converge fast and yield less offset when the
meteorological data are applied.

Figure C. 1 Pseudorange performance [WES2, DOY136 (top) and 137 (bottom)(2000)]

263
Appendix D: HOURLY JUMPS IN RINEX FILES

It is found that some pseudorange observations, but not carrier phase, are adjusted hourly for
the receiver clock error. Examples include RINEX observations from Ashtech and Trimble
receivers. Figure D.1 shows the sum of the L1 and L2 observations. These hourly jumps are
1 ms or multiple of 1 ms and must be corrected before observations can be subjected to cycle
slip detection and fixing.

Figure D. 1 Hourly jumps in pseudorange P1+P2 (top), but not carrier phase L1 + L2

(bottom) for PRN4 [GAIT, DOY2(2000)]

264
BIOGRAPHY OF THE AUTHOR

Boonsap Witchayangkoon was born and raised in Songkhla, Thailand. He graduated from

The Mahavajiravudh High School in 1988. He received his Bachelor of Engineering (Civil

Engineering) degree with Honors from The King Mongkut's University of Technology

Thonburi (previously known as The King Mongkut's Institute of Technology Thonburi),

Bangkok, Thailand, in 1992. He worked at construction consulting companies as a civil

engineer and at universities as a lecturer.

In 1994, he won the Royal Government Scholarship awarded for higher study in USA.

In 1995, he enrolled for graduate studies in Spatial Information Science and Engineering at the

University of Maine. He earned his Master's degree in 1997. He will be working in the

Department of Civil Engineering, Thammasat University, Pathumtani, 12121, Thailand. He is

a candidate for the Doctor of Philosophy degree in Spatial Information Science and

Engineering from The University of Maine in December, 2000.

265

You might also like