You are on page 1of 60

An

Analysis of Ocean Tide Loading and


Modelling Strategies for GNSS
Precise Point Positioning in Britain
& Ireland

Peadar Quine

BSc (Honours) Geomatics (Surveying & Mapping)


Table of Contents
List of Figures........................................................................................................................ 4
List of Tables .........................................................................................................................4
List of Equations.................................................................................................................... 4
Acknowledgements................................................................................................................4
List of Acronyms....................................................................................................................5
1Introduction.......................................................................................................................... 7
1.1Project Goals..................................................................................................................7
2Literature Review.................................................................................................................8
2.1Overview of GNSS........................................................................................................8
2.1.1GNSS Signal Structure............................................................................................9
2.1.2GNSS Positioning Fundamentals ...........................................................................9
2.1.3Clock Errors.......................................................................................................... 10
2.1.4Orbital Errors........................................................................................................ 11
2.1.5Atmospheric Errors...............................................................................................11
2.2Differential GNSS....................................................................................................... 11
2.2.1Network Real-Time Kinematic GNSS..................................................................12
2.3Precise Point Positioning............................................................................................. 12
2.3.1International GNSS Service .................................................................................13
2.3.2PPP Algorithms.....................................................................................................14
2.4PPP Online Post-Processing Services..........................................................................16
2.4.1CSRS-PPP Service................................................................................................17
2.4.2Trimble RTX-PP...................................................................................................17
2.5PPP Desktop Software.................................................................................................18
2.5.1GIPSY-OASIS...................................................................................................... 18
2.5.2gLAB.....................................................................................................................18
2.5.3RTKLIB................................................................................................................ 18
2.6Ocean Tide Loading ................................................................................................... 19
2.7The International Terrestrial Reference Frame ...........................................................23
2.8Vertical Reference Systems.........................................................................................23
2.8.1The Geoid, Ellipsoid and Mean Sea Level .......................................................... 24
2.9Python..........................................................................................................................24

2
3Methodology...................................................................................................................... 26
3.1Coordinate Reference System......................................................................................27
3.2Site Selection & Data Acquisition...............................................................................27
3.2.1British Isles Continuous GNSS Facility ...............................................................28
3.2.2The UK Tide Gauge Network ..............................................................................28
3.2.3The Irish National Tide Gauge Network .............................................................28
3.2.4Sites.......................................................................................................................28
3.2.5Tidal Data..............................................................................................................30
3.2.6GNSS Data............................................................................................................30
3.3Baseline Elevation.......................................................................................................30
3.4OTL Files Generation..................................................................................................31
3.5Time-Window for GNSS Processing...........................................................................31
3.6CSRS PPP Processing..................................................................................................31
3.6.1TEQC.................................................................................................................... 32
3.6.2Ocean Tide Loading Parameters...........................................................................32
3.6.3Data Extraction..................................................................................................... 33
3.7RTKLIB Processing.....................................................................................................34
3.7.1RTKPOST Settings...............................................................................................35
3.7.2Data Extraction..................................................................................................... 38
3.8Tidal Data Processing..................................................................................................39
3.8.1Tidal Phase Shift................................................................................................... 40
3.9Collation of Processing Results...................................................................................40
4Analysis and Results.......................................................................................................... 42
4.1CSRS-PPP versus RTKPOST Solutions..................................................................... 43
4.1.1SCIL: RTKPOST No OTL v CSRS-PPP No OTL...............................................43
4.1.2SCIL: RTKPOST GOT00.2_Solid v CSRS-PPP GOT00.2_Solid.......................45
4.1.3SCIL: RTKPOST GOT00.2_CMM v CSRS-PPP GOT00.2_CMM.................... 47
4.1.4BELF: RTKPOST No OTL v CSRS-PPP No OTL..............................................48
4.1.5BELF: RTKPOST GOT4.7 v CSRS-PPP GOT4.7...............................................50
4.1.6Summary of Findings............................................................................................52
5Discussion.......................................................................................................................... 54
5.1Recommendations for Future Studies..........................................................................54
6Conclusion..........................................................................................................................55

3
References............................................................................................................................56

List of Figures

List of Tables

List of Equations

Acknowledgements
BIGF

BODC

Marine Institute

4
List of Acronyms
BIGF British Isles Continuous GNSS Facility

CMM Centre of Mass Motion

CORS Continually Operating Reference Stations

CSRS-PPP Canadian Spatial Reference System PPP

DGNSS Differential GNSS

DOP Dissolution of Precision

DORIS Doppler Orbitography and Radio-Positioning Integrated by Satellite

EOP Earth Orientation Parameters

ERP Earth Rotation Parameters

ESA European Space Agency

gAGE Research Group of Astronomy and Geomatics

GDOP Geometric DOP

GIPSY-OASIS GNSS-Inferred Positioning System and Orbit Simulation Software

gLAB The GNSS-LAB Tool

GLONASS Global Orbiting Navigation Satellite System

GNSS Global Navigation Satellite Systems

GOTxx Goddard Ocean Tide Model

GPS Global Positioning System

GRS80 Geodetic Reference System 1980

GUI Graphic User Interface

ICRF International Celestial Reference Frame

IERS Earth Rotation and Reference Systems Service

IGS International GNSS Service

ITGN The Irish National Tide Gauge Network

ITRF International Terrestrial Reference Frame

5
ITRS International Terrestrial Reference System

LLR Lunar Laser Ranging

LPSNI Land and Property Service of Northern Ireland

MSL Mean Sea Level

NRTK Network Real-Time Kinematic GNSS

OSGB Ordnance Survey of Great Britain

OSi Ordnance Survey of Ireland

OTL Ocean Tide Loading

PCV Phase Centre Variation

PPP Precise Point Positioning

PRN Pseudo-Random Noise

RINEX Receiver Independent Exchange Format

RTK Real-Time Kinematic

RT-PPP Real-Time PPP

UKTGN The UK Tide Gauge Network

VLBI Very Long Baseline Interferometry

6
Introduction

1 Introduction

1.1 Project Goals


1. Develop suitable Python scripts for automated manipulation of large volumes of
CSRS-PPP service.

2. Develop suitable Python scripts for automated manipulation of large volumes of


RTKLIB post-processing results.

3. Develop an appropriate RTKLIB PPP processing configuration for the generation


of PPP solutions with comparable accuracy to existing online services.

4. Assess the impact of OTL on PPP solutions in coastal areas of Britain and Ireland.

5. Assess the ability of OTL files to mitigate the impact of OTL in coastal areas of
Britain and Ireland.

6. Assess if there is a significant difference between using GOT00.2 OTL models


referenced to the Solid Earth against those corrected for Centre of Mass
Motion" in PPP processing.

7. Assess if there is a significant difference between PPP results generated in


RTKLIB using the GOT00.2 and GOT4.7 OTL models.

8. Assess the phase difference between tidal cycles and OTL in Britain and Ireland.

7
Literature Review

2 Literature Review
Precise point positioning (PPP) is an evolving GNSS solution which can produce sub-
decimetre positional accuracies (Throux and Kouba 2001). This is achieved through
the use of International GNSS Service (IGS) products and requires enhanced bias
modelling (Grinter and Janssen 2012). Ocean loading displacement is among the
various error sources that could impact on GNSS positioning solutions. The requirement
for implementation of ocean tide loading (OTL) mitigation strategies in the various PPP
solutions currently available is not consistent and is often considered a matter of
application or location (Natural Resources Canada 2014). This literature review will
attempt to place the PPP GNSS technique in context, and discuss existing software,
services and products required for successful implementation of PPP algorithms. The
review will also examine the significance of various bias sources with particular
emphasis on OTL.

2.1 Overview of GNSS


Global Navigation Satellite Systems (GNSS) is the term commonly used to describe all
available global satellite positioning systems and incorporates the commonly used U.S
militarys Global Positioning System (GPS) and Russias Global Orbiting Navigation
Satellite System (GLONASS), amongst others. A GNSS is commonly viewed as having
three distinct components; the space segment, the control segment and the user segment.
The space segment refers to the satellite constellations and the broadcast signals which
identify each satellite, the time, orbit and the satellites health status (Jeffrey 2010). The
control segment refers to the ground based satellite monitoring stations which send
predicted ephemeris corrections to the satellites which are in turn included in the
satellites broadcast signal. The user segment refers to the various end user receivers,
each receivers ability to interpret different GNSS signals and the associated
applications (Jeffrey 2010; Ghilani and Wolf 2012; Subirana, Zornoza and Hernndez-
Pajares 2013). The concept of the three GNSS segments is illustrated in Figure 2.1.

Figure 2.1 GNSS Segments (Jeffrey 2010)

8
Literature Review

2.1.1 GNSS Signal Structure


GNSS satellites transmit electromagnetic signals using the L-Band. These sinusoidal
carrier signals carry the systems ranging codes. The ranging codes are binary codes
(sequences of zeros and ones) which can be used by the receiver to identify the time lag
between propagation and reception (Subirana, Zornoza and Hernndez-Pajares 2013).
These codes are commonly referred to as Pseudo-Random Noise (PRN) sequences
(Strang and Borre 1997; Jeffrey 2010; Subirana, Zornoza and Hernndez-Pajares 2013).
Further binary codes that contain navigational data such as ephemeris and clock
information are commonly referred to as the navigation message.

2.1.1.1 GPS Signals


GPS satellites all use the same carrier phase frequencies and individual satellites are
identified through unique modulated PRN sequences of 1023 bits. Initially two binary
ranging codes were transmitted in GPS, a Coarse/Acquisition (C/A) code with a rate of
1.023 MHz; and an encrypted Precision (P(Y)) code which is approximately 10 times
faster and was restricted to military use (Strang and Borre 1997). Prior to December
2005 the C/A code was broadcast on both the L1 (1575.42 MHz) and L2 (1227.60
MHz) carrier frequencies and the encrypted P(Y) code was only broadcast on the L2
band (ESA 2014).

Since 2005 the US government has developed a modernisation strategy to improve the
quality of GPS (ESA 2014). This strategy introduced three new civilian ranging signals,
L1 Civilian (L1C), L2 Civilian Moderate (L2CM) and L2 Civilian Long (L2CL). L1C is
carried on the L1 band and the L2 band carries the L2CM and L2CL codes. A
modernised military code (M-code) is carried on both L1 and L2. In addition a new
carrier frequency, L5, was introduced in 2010. The L5 frequency carries two PRN
ranging codes, I5 and Q5, at 1176.45 MHz (ESA 2011d; 2014). Table 2.1 provides a
summary of the GPS carrier signals and PRN ranging codes.

Table 2.1 GPS Carrier Signals and PRN Ranging Codes (ESA 2014)

Carrier Signal Frequency PRN Codes


L1 1575.42 MHz C/A, L1C, P(Y), M-code
L2 1227.60 MHz C/A, L2CM, L2CL, P(Y), M-code
L5 1176.45 MHz L5I, L5Q

2.1.1.2 GLONASS
GLONASS signal structure differs from GPS in that each satellite propagates similar
PRN ranging codes on sub-frequencies within the L1 and L2 bands (Russian Institute of
Space Device Engineering 2008). It is through identification of these sub-frequencies
that different satellites are identified in GLONASS (ESA 2011c).

2.1.2 GNSS Positioning Fundamentals


The basic positioning technique common to all GNSS solutions is the geometric
computation of a receiver position from the distances determined from satellites of
known position (Subirana, Zornoza and Hernndez-Pajares 2013). The distance from
satellite to receiver is determined by observing the time taken for an electromagnetic

9
Literature Review

signal propagated from the satellite to be received by the user. This travel time
multiplied by the speed of light provides a measurement of the distance (Subirana,
Zornoza and Hernndez-Pajares 2013). This principal equation is:
(Eq. 0)

Where:
d = the distance travelled
c = the mean speed of light along the propagated signals trajectory
t = the time lag between propagation and reception
(Subirana, Zornoza and Hernndez-Pajares 2013)

The preceding equation represents the theoretical situation in which observations are
free from systematic errors. In practice this situation is unlikely. Therefore an error of
unknown magnitude r is introduced as:
(Eq. 0)

(Subirana, Zornoza and Hernndez-Pajares 2013)

Due to the uncertainty associated with these errors the initial observations are
commonly referred to as pseudo-ranges (Jeffrey 2010; Ghilani and Wolf 2012;
Subirana, Zornoza and Hernndez-Pajares 2013). Given sufficient pseudo-ranges are
observed and the positions of the satellites are known; trilateration calculations can be
used to solve for the three positional unknowns x, y and z (Strang and Borre 1997).

2.1.3 Clock Errors


As previously suggested (Section GNSS Positioning Fundamentals) the key observable
in GNSS positioning is time. To mitigate potential timing errors GNSS satellites are
equipped with highly accurate atomic oscillators. These atomic clocks are a critical
component of any GNSS (Subirana, Zornoza and Hernndez-Pajares 2013). Although
highly accurate the satellite clocks can accumulate small timing offsets. To mitigate this
they are continuously monitored by the ground segment and clock corrections are
transmitted to the satellites and forwarded to the user segment through the navigation
message (Jeffrey 2010).

Due to the expense associated with atomic clocks it is not feasible to have such clocks
built into user receivers. Therefore GNSS receivers utilise relatively inexpensive quartz
clocks. These quartz clocks are accurate to approximately 5 parts per million which can
introduce range errors in the region of 1500m (Jeffrey 2010). As the same receiver
clock is used, the timing offset is considered constant between pseudo-range
observations. To solve these receiver timing errors the time offset is introduced to the
positional solution as a fourth unknown (Subirana, Zornoza and Hernndez-Pajares
2013). This means that a minimum of four satellites are required to solve the four
unknowns (x, y, z and t). The incorporation of additional satellites will introduce
observation redundancy and strengthen the solution (Jeffrey 2010).

10
Literature Review

2.1.4 Orbital Errors


Precise knowledge of satellite orbits is a vital component of GNSS positioning. GNSS
satellites include ephemeris information in their navigation signals (Subirana, Zornoza
and Hernndez-Pajares 2013). Orbital errors refer to discrepancies between the
broadcast position and velocity of a satellite and the satellites true position and velocity
(Colombo 1986). Satellite orbits can be accurately approximated using algorithms based
on celestial mechanics which provide Keplerian orbits (Subirana, Zornoza and
Hernndez-Pajares 2013).

Harmonic modelling is used to compensate for orbital perturbations associated with the
Earths non-homogenous gravitational potential. The gravitational influence of other
celestial bodies, particularly the Sun and Moon are also considered. Solar radiation
pressure also influences a perturbing force on satellite orbits and is particularly difficult
to model. The combined effect of these orbital influences can become significant over
time (Subirana, Zornoza and Hernndez-Pajares 2013).

2.1.5 Atmospheric Errors


As the velocity and phase of electromagnetic waves are affected by the median through
which they pass, the effects of the Earths atmosphere on propagated GNSS signals
must be mitigated for (Leick 2004). Two distinct regions of the Earths atmosphere are
generally considered in GNSS applications, the troposphere and the ionosphere.

The troposphere is the region of the Earths atmosphere which contains the majority of
the Earths atmospheric mass. The troposphere has an effective height of approximately
40km above the Earths surface. The hydrostatic component of the troposphere can
produce an error of approximately 2.4m at sea level whilst the non-hydrostatic wet
component is responsible for up to 0.4m error in pseudo-range observations (Leick
2004). In GNSS the effects of the troposphere are generally mitigated through the use of
modelling and mapping techniques.

The ionosphere represents the upper layers of the Earths atmosphere to approximately
2000km above the Earths surface. Ionospheric delay in GNSS is dependent on the
variable electron density within the ionosphere (Subirana, Zornoza and Hernndez-
Pajares 2013). Variations within the ionosphere will also cause GNSS code and signal
phase shifts (Leick 2004). These are mitigated in single frequency GNSS through
ionospheric models which are broadcast with the GNSS navigation signal. In multi-
frequency GNSS the interdependency between ionospheric errors and electromagnetic
signal frequency facilitates the elimination of the error using ionospheric-free functions
(Leick 2004; Subirana, Zornoza and Hernndez-Pajares 2013).

2.2 Differential GNSS


Differential GNSS (DGNSS) is a widely used technique that mitigates many of the
ranging errors associated with GNSS (Jeffrey 2010). DGNSS techniques utilise one or
more known reference stations that are simultaneously acquiring GNSS positions. The
measured deviation from the reference stations and the required corrections to the
pseudo-ranges can then be applied to unknown receiver positions (ESA 2011b). The

11
Literature Review

technique assumes that the ephemeris, satellite clock and atmospheric pseudo-range
errors do not vary excessively over time and space (ESA 2011a). The correction signals
can be broadcast and received in real-time to provide instantaneous solutions or can be
applied in post-processing. High precision DGNSS positioning can be achieved through
the use of carrier phase signals in addition to satellite navigation code (Weston and
Schieger 2010). The basic principles of DGNSS are illustrated in Figure 2.2.

Figure 2.2 Differential GNSS Schematic (Jeffrey 2010)

2.2.1 Network Real-Time Kinematic GNSS


The basic concept of DGNSS has been developed to produce a range of solutions.
Arguably the most significant of these developments for the generation of real-time,
high precision GNSS positioning is Network Real-Time Kinematic (NRTK)
positioning. This solution utilises a network of reference stations providing a robust
solution that mitigates the spatial decorrelation in pseudo-range error sources associated
with single station DGNSS techniques and introduces redundancy to the solution (El-
Mowafy 2012). Networks of continually operating reference stations (CORS) have been
established in most countries with a developed survey infrastructure and the NRTK
technique has become the principal technique for the collection of high precision
positional data in these countries.

2.3 Precise Point Positioning


Precise Point Positioning (PPP) offers an alternative GNSS precise positioning solution
to the range of widely used DGNSS techniques. PPP utilises undifferenced pseudo-
range and carrier phase signals (Throux and Kouba 2001), a single GNSS receiver and
precise satellite ephemeris and clock corrections (Rizos et al. 2012). There are a range
of PPP algorithms currently available that can provide cm to decimetre accuracy (Van
De Marel and De Baker 2012).

12
Literature Review

The principal advantage of PPP over the various DGNSS solutions is the users
requirement for only a single GNSS receiver. This reduces costs and enables the
technique to be used in regions that do not have the necessary survey infrastructure for
the various differential techniques (Grinter and Janssen 2012). PPP is also advantageous
in that it uses products that are directly referenced to a global reference frame. This is a
significant advantage over the differential approach in which solutions are determined
relative to the local reference station or stations (Gao 2006; ESA 2011h).

Although PPP can be undertaken using either single or dual-frequency receivers (Van
De Marel and De Baker 2012; Subirana, Zornoza and Hernndez-Pajares 2013) the use
of undifferenced carrier phase and pseudo-range measurements restricts the PPP
approaches ability to mitigate atmospheric biases in single-frequency solutions (Rizos
et al. 2012). Ionospheric-free functions are however available in dual-frequency PPP
solutions. As PPP utilises undifferenced GNSS observations a range of additional bias
sources that are not normally considered in differential techniques must be considered.
These include phase wind-up corrections, satellite antenna phase centre corrections and
site displacement corrections (Grinter and Janssen 2012).

PPP solutions have convergence times in excess of 20 minutes which has inhibited their
use in applications that require real-time solutions (Grinter and Janssen 2012). However
real-time PPP (RT-PPP) solutions that provide decimetre accuracy are available and
include products such as Fugros Starfix.XP2, Starfix.G2 and Starfix.G4 services
(Sagatun 2015). These commercial solutions utilise private satellite tracking stations
and proprietary algorithms. However the recent introduction of a real-time delivery
service for precise ephemeris and clock products and scientific GNSS processing
software has enabled the demonstration of decimetre level RT-PPP kinematic solutions
for non-proprietary users (Rizos et al. 2012).

2.3.1 International GNSS Service


The International GNSS Service (IGS) is a voluntary federation of over 200 agencies in
over 100 countries. The IGS operates a global network of over 400 GNSS tracking
stations and from these compiles a wide range of GNSS products. IGS products include
Earth rotation parameters, tropospheric path delay estimates and ionospheric maps (IGS
2015a). The IGS also provides the sub-decimetre accuracy GNSS orbit information and
sub-nanosecond accurate satellite clock information required for high accuracy PPP
GNSS solutions (Throux and Kouba 2001). All IGS products are referenced directly
to the International Terrestrial Reference Frame (ITRF). Use of IGS products in PPP
eliminates the requirement for sophisticated global analysis as this analysis has already
been undertaken within the IGS Analysis Centres before publication (Kouba 2009).

IGS products are generally divided into three forms based on the latency of the product.
These distinctions are the Ultra-Rapid, Rapid and Final products (Kouba 2009). The
accuracy and latency of the IGS clock and ephemeris products are illustrated in Table
2.2. The accuracies of the GNSS broadcast clock and ephemerides, which are not IGS
products, are included for comparison purposes.
Table 2.2 IGS Clock and Ephemeris Product Accuracy and Latency (IGS 2015b)

13
Literature Review

Type Orbit Satellite Clocks


RMS Interval RMS Interval
Broadcast ~100cm Daily ~5ns Daily
Ultra-Rapid ~5cm 15min ~3ns 15min
(predicted half)
Ultra-Rapid ~3cm 15min ~150ps 15min
(observed half)
~2.5cm 15min ~75ps 5min
Rapid
Final ~2.5cm 15 min ~75ps 30s

2.3.2 PPP Algorithms


Dual frequency PPP algorithms utilise undifferenced carrier phase and pseudo-range
observations in conjunction with satellite ephemeris and clock corrections. The
simplified observation equations for each satellite at a single epoch can be written as:
(Eq. 0)

(Eq. 0)

Where:
= the ionosphere-free combination of L1 and L2 pseudo-ranges
= the ionosphere-free combination of L1 and L2 carrier phases
= the receiver clock offset from GPS time
= the satellite clock offset from GPS time
c = the speed of light in a vacuum

Tr = the tropospheric path delay

= the carrier combination wavelength

N = the ambiguity of the carrier-phase ionosphere-free combination

14
Literature Review

/ = the observation noise components

= the range between the satellite and the receiver, computed as a function of the

satellite (x , y , z ) and receiver (x , y , z ) coordinates as:


Sat Sat Sat Rx Rx Rx

(Eq. 0)

(ESA 2011e)

The tropospheric delay (Tr) can then be expressed as a function of the zenith path delay

(zpd) and mapping function (M). The satellite clock offsets () are known from the IGS
clock corrections and can be removed. The equations can then be simplified to the form:

- (Eq. 0)

- (Eq. 0)

(Kouba 2009)

The observation equations can then be linearized to matrix form and unknowns solved
for using least squares adjustment (Kouba and Theroux 2000). Combining the
observations from all satellites at a single epoch in the matrix enables the unknowns in
the equation (e.g. the receiver coordinates, the receiver clocks, phase ambiguities and
tropospheric delay) to be solved using filters (ESA 2011e). Generally sequential filters
that facilitate dynamic inputs are used in the adjustment procedure (Kouba and Theroux
2000).

For high accuracy PPP additional biases that are not normally required for most
differential positioning applications must be considered (Kouba 2009). Table 2.3 details
and compares the various corrections and error models that are commonly applied to
dual-frequency PPP solutions and high accuracy differential techniques.

15
Literature Review

Table 2.3 Dual-Frequency PPP and DGNSS Solution Corrections (Grinter and Roberts 2011)

Correction/Model PPP DGNSS


Precise Satellite Clock Corrections
Satellite Antenna Phase Centre Offset
Satellite Antenna Phase Centre Variations
Precise Satellite Orbits /
Satellite Antenna Phase Windup Error
Receiver Antenna Phase Centre Offset
Receiver Antenna Phase Centre Variations
Receiver Antenna Phase Windup
Site Displacement Models /
Earth Tides only
Tropospheric Modelling

Provide description of the Corrections

The number of satellites and the quality of the geometry between them at any given
epoch will have a significant influence on the accuracy of the PPP solution. The quality
of the signals with respect to noise factors such as multipath will also impact on the
final solution (ESA 2011e). For this reason PPP algorithms can include additional
parameters such as elevation masks and dissolution of precision (DOP) rejection
thresholds to ensure that a minimum observation quality is maintained.

2.4 PPP Online Post-Processing Services


Online PPP post-processing services have been developed by various public and private
organisations over the last number of years (ESA 2011h). In general they provide an
easy to use, cost free services that facilitate worldwide cm level positioning (Silver
2013). Available services include:

Online Positioning User Service (OPUS)

AUSPOS

GPS Analysis and Positioning Software (GAPS)

Magic GNSS

Automatic Precise Positioning Service (APPS)

Canadian Spatial Reference System PPP (CSRS-PPP)

TrimbleRTX-PP

OMahony (2014) undertook a study of the accuracies of some of the listed PPP
services for a station position in Ireland using different receivers over several time

16
Literature Review

periods and found that all the services investigated provided comparable results. A
summary of OMahonys findings is contained in Table 2.4.
Table 2.4 Summary of Findings from PPP Service Accuracy Analysis (O'Mahony 2014)

Horizontal
Rank Service Vertical Accuracy
Accuracy
1 Trimble RTX +/-0.002m +/-0.004m
2 CSRS-PPP +/-0.003m +/-0.016m
3 OPUS +/-0.004m +/-0.024m
4 AUSPOS +/-0.008m +/-0.023m
5 Magic GNSS +/-0.010m +/-0.030m

This study will focus on the use of the CSRS-PPP service and will utilise Trimble RTX
for the establishment of baseline positions for analysis and comparisons.

2.4.1 CSRS-PPP Service


CSRS-PPP was developed by Natural Resources Canada (NRCan). CSRS-PPP is
capable of processing both GPS and GLONASS satellite observations and facilitates the
processing of dual or single frequency observation files using either static or kinematic
modes. Submitted data must be in Receiver Independent Exchange Format (RINEX) or
Compact RINEX format and results are returned to the user by email (Natural
Resources Canada 2014).

CSRS-PPP provides a graphic rich PDF report of the processed results, CSV and POS
files containing epoch by epoch positions and a SUM file with details of the algorithm
parameters. The three latter formats are all easily read and manipulated with automated
scripting methods. CSRS-PPP also facilitates the inclusion of OTL files at the users
discretion. It is the availability of machine readable files and the ability to process data
with and without ocean loading files that encouraged the use of CSRS-PPP for this
study. The service also allows several RINEX files to be zipped together and uploaded
simultaneously and will provide separate results for each file. CSRS-PPP processes both
GPS and GLONASS satellites in the solution.

2.4.2 Trimble RTX-PP


TrimbleRTX-PP is a service that merges traditional PPP with elements of NRTK (Rizos
et al. 2012). The service was released in 2011 and provides a real-time positioning
service via satellite or internet links. The service utilises a global network of satellite
stations to provide satellite ephemeris and clock corrections (Doucet et al. 2012). The
TrimbleRTX-PP service is a web-based service that allows post-processing of GNSS
data in RINEX or Trimble proprietary formats. Unlike traditional PPP services,
TrimbleRTX-PP utilises networks of regional CORS stations to determine local
atmospheric corrections (Rizos et al. 2012). Trimble RTX-PP can only be used with
compatible GNSS receivers and currently includes processing of GPS, GLONASS,
QZSS and BeiDou satellite services (Trimble 2015). Trimble RTX-PP utilises the
GOT00.2 tidal model (Ahmed 2014).

17
Literature Review

The output from the Trimble RTX-PP service is a PDF report emailed to the user. The
service allows for multiple observation files for the same receiver to be zipped together
and submitted as long as the observations do not exceed 24 hours. Multiple submissions
are however combined to produce a single position. This restricted the ability of
Trimble RTX-PP to be used to process the large number of files required for this study
and so the service was used only to provide baseline data for each station.

2.5 PPP Desktop Software


There are several desktop applications currently available that are capable of providing
PPP solutions; although there are currently no proprietary desktop packages with the
capability (Rizos et al. 2012). Some of the available programs are summarised below.

2.5.1 GIPSY-OASIS
The GNSS-Inferred Positioning System and Orbit Simulation Software GIPSY-OASIS)
is software developed by NASAs Jet Propulsion Laboratory. The software is only
available under licence from the Jet Propulsion Laboratory and was not available for
this study.

2.5.2 gLAB
The GNSS-LAB Tool (gLAB) was developed by the Research Group of Astronomy and
Geomatics (gAGE) as an educational tool for the European Space Agency (ESA) and
facilitates the post-processing of standalone GNSS and PPP solutions. The gLAB
program facilitates most modelling and corrections normally required for accurate PPP;
however OTL is merely considered as a tertiary effect in gLAB and cannot be modelled
(Subirana, Zornoza and Hernndez-Pajares 2013). Because of the lack of OTL
modelling capabilities the gLAB tool was not utilised in this study.

2.5.3 RTKLIB
RTKLIB (Takasu 2007) is an open source desktop software package that provides a
range of GNSS processing capabilities including real-time kinematic (RTK) and PPP
solutions. RTKLIB comprises of a portable program library and several desktop
applications with graphic user interfaces (GUIs) (Takusa 2013). The program was
developed in 2006 and has had a number of releases since this date. The version used in
this study was version 2.4.3 which at the time of writing was available in beta form.

For PPP solutions with RTKLIB the RTKGET and RTKPOST applications are utilised.
RTKGET provides an easy to use application that will automatically download satellite
clock and ephemeris correction files for selected dates. The RTKGET GUI (Figure 2.3)
is also capable of retrieving other published IGS products such as earth rotation
parameters (ERPs), antenna correction files (ANTEX Files), phase centre variation files
(PCVs) and ionospheric TEC grids. Products similar to the IGS products from other
sources are also available.

18
Literature Review

Figure 2.3 RTKGET GUI

RTKPOST is the post-processing application of RTKLIB. It facilitates post-processing


of RTK and PPP GNSS from RINEX format observation files and associated navigation
message files. PPP solutions can be processed in Static or Kinematic mode and can use
GPS, GLONASS, BeiDou and QZSS and Galileo satellite constellations (Takusa 2013).
RTKPOST facilitates the time-windowing of RINEX files during processing to produce
results for the selected windows and allows optimal user control for the implementation
of PPP corrections and modelling. The GUI allows the user to set statistical parameters,
satellite elevation masks and geometric DOP (GDOP) thresholds. The

2.6 Ocean Tide Loading


Ocean tide loading (OTL) can be described as the displacement of the Earths crust due
to the periodic tidal redistribution of ocean mass. OTL represents a substantial GNSS
error source as it can cause a significant displacement in a sites mean position over the
space of a few hours (Clarke and Penna 2010). The principal impact of OTL is on the
vertical component of any observation with the impact on horizontal components in the
region of a third of the vertical displacement (Clarke and Penna 2010).

The ocean tides are principally influenced by the gravitational interrelationships


between the Earth, Moon and Sun. Due to the different gravitational forces that
influence the ocean tides, OTL can be considered as the combination of several ocean
tides, each with independent periods (Bos and Scherneck 2015). In addition to these
astronomical influences, the ocean tides are subject to the influences of hydraulic,
hydrodynamic, hydrographic, topographic and meteorological conditions (Center for
Operational Oceanographic Products and Services 2013).

OTL is dominated by diurnal and semi-diurnal periods in much the same way as solid
Earth tides (Kouba 2009). Modelling of the ocean tides is undertaken using well
established harmonic constituents and harmonic analysis techniques. A simple harmonic
equation is demonstrated in (Eq. 8) below.

19
Literature Review

X(t) = Hn cos (n t gn ) (Eq. 0)


Where:

X = Value of the variable quantity at time t

H = Amplitude of oscillation
n

g = Phase lag
n

= Angular speed
n

t = Time

(UK Hydrographic Office 2006)

OTL displacements are modelled through the use of ocean tide models and an Earth
model that describes the Earths response to a unit load (Clarke and Penna 2010). Ocean
tide models are generated with the aid of satellite altimetry missions such as the
Topex/Poseiden mission, the Jason-1 mission and the more recent Ocean Surface
Topography Mission (OSTM)/Jason-2 mission. The quality and accuracy of the tidal
models would be expected to improve with the availability of extended time series data
for the phase and amplitude of the tides and also with advances in data assimilation
techniques (Shum et al. 1997). The altimetry data used in the tidal models is often
supplemented and verified using global tide gauge networks. Models such as the
Goddard Ocean Tide model (GOTxx) are also supplemented with the use of existing
hydrodynamic models such as the Finite Element Solution (FES) (Ray 1999). This
study will attempt to assess if there is a significant difference between using the
GOT00.2 derived OTL model and the more recent GOT4.7 model.

Normally 11 harmonic constituents are considered in OTL modelling (Bos and


Scherneck 2015). The tide is principally influenced by four semi-diurnal constituents
and four diurnal constituents. The remaining constituents are fortnightly, monthly and
semi-annual (Clarke and Penna 2010). The fortnightly harmonics are commonly
referred to as the spring and neap tides and the diurnal and semi-diurnal constituents
combine to provide the familiar twice daily (semi-diurnal) high and low tidal patterns
that occur in Britain and Ireland.

20
Literature Review

Table 2.5 illustrates the 11 principal harmonic constituents and their associated periods.
The global tidal amplitude has been normalised to the most significant constituent, M2,
for illustrative purposes. Tidal amplitude is latitude dependent and is also affected by
various non-astronomical influences (Clarke and Penna 2010).
Table 2.5 Principal Tidal Constituents, Periods and Tidal Potential Normalised to the M2 Harmonic (Clarke
and Penna 2010)

Normalised
Constituent Description Period
Potential
M2 Principal lunar semidiurnal constituent 12 h 25 min 1.00
S2 Principal solar semidiurnal constituent 12 h 00 min 0.47
Larger lunar elliptic semidiurnal
N2 12 h 39 min 0.19
constituent
K2 Lunisolar semidiurnal 11h 58 min 0.13
K1 Lunar diurnal constituent 23 h 56 min 0.58
O1 Lunar diurnal constituent 25 h 50 min 0.41
P1 Solar diurnal constituent 24 h 04 min 0.19
Larger lunar elliptic diurnal
Q1 26 h 52 min 0.08
constituent
Mf Lunisolar fortnightly constituent 13.66 days 0.11
Mm Lunar monthly constituent 27.32 days 0.06
Ssa Larger lunar evectional constituent 182.62 days 0.05

OTL models can be generated using the Onsala Space Observatorys Free Ocean Tide
Loading Provider (FLP) site (http://holt.oso.chalmers.se/loading/) which uses
OLFP/OLMG software (Scherneck 1991). The service enables a range of OTL models
to be generated based on the various tidal models and uses the OLMP. The system also
enables the OTL models to be generated using the solid Earth as a reference or allows
for a centre of mass motion (CMM) correction. This concept refers to the periodic
movement of the Earths centre of mass due to the ocean and Earth tides. This
movement may be detectable using GNSS or may be absorbed in the GNSS ephemeris
observations (Scherneck 2011). The service does not provide a recommendation as to
whether the solid Earth or the CMM versions of the models are best suited for use in
PPP solutions and it appears no recent literature on the matter has been produced.
Section 2.5 of the CSRS PPP .SUM files that are generated as solution outputs suggest
that the OTL model utilised should use the solid Earth as there centre (Figure 2.4). In an
attempt to clarify this issue this study will include a comparison of the solid Earth and
CMM model types using the GOT00.2 loading model.

Figure 2.4 Extract from a CSRS PPP .SUM File

21
Literature Review

Due to global variations in tidal magnitude, the magnitude of ocean tide loading
displacement is also variable. Because of unique seafloor topography, southwest
Ireland and Britain have one of the highest global OTL displacements; with the peak-to-
peak OTL displacement reaching up to 13cm (Clarke and Penna 2010). Figure 2.5
illustrates the global variation in OTL displacement as generated from the TPXO6.2 tide
model (Egbert and Erofeeva 2002).

Figure 2.5 Global OTL Displacement using TPXO6.2 Tidal Model (Egbert and Erofeeva 2002)

The necessity for implementing an OTL model in PPP algorithms is generally


considered as being a matter of the required accuracy, observation period and location
(Natural Resources Canada 2014). The developers of the gLAB software package
considered OTL as a secondary site displacement effect and did not include OTL
modelling capability in the software (Subirana, Zornoza and Hernndez-Pajares 2013).
Clarke & Penna (2010) predicted that NRTK positioning can, in inland areas of Britain
and Ireland, successfully mitigate much of the OTL error without requiring additional
modelling. However Clarke & Penna (2010) also identified that in some coastal regions
of Britain and Ireland residual OTL errors of up to 1cm may remain in post-processed
NRTK solutions. No examination of the impact of OTL displacement on PPP solutions
in Ireland and Britain has yet been published. It is considered that OTL displacement is
likely to be insignificant in 24 hour PPP observations or in observations at a distance
greater than 1000km from the coast but that it should be considered in static PPP
positioning at the cm level when observation periods are considerably less than 24 hours
(Throux and Kouba 2001). It is possible that the magnitude of the OTL displacement
may not be sufficient magnitude to be clearly identified and will be masked by the noise
inherent in the PPP solution. This study aims to clarify the significance of OTL in
Ireland and Britain with respect to PPP GNSS solutions.

22
Literature Review

2.7 The International Terrestrial Reference Frame


The International Terrestrial Reference Frame (ITRF) is maintained by the International
Earth Rotation and Reference Systems Service (IERS) and is linked to International
Celestial Reference Frame (ICRF) by the Earth Orientation Parameters (EOP). ITRF at
the various epochs represent realisations of the dynamic International Terrestrial
Reference System (ITRS) and are published with velocities for the various sites that
form the reference frame and transformation parameters to other ITRF solutions (IGN
2015b).

ITRF is realised through the use of IERS reference stations which incorporate data from
Doppler Orbitography and Radio-Positioning Integrated by Satellite (DORIS), GNSS,
Lunar Laser Ranging (LLR), Satellite Laser ranging (SLR) and Very Long Baseline
Interferometry (VLBI) techniques (Altamimi, Collilieux and Mtivier 2011). The
locations of the reference stations used in the ITRF 2008 realisation are illustrated in
Figure 2.6. The most recent realisation is ITRF2008 with a new realisation, ITRF2013,
currently in progress.

Figure 2.6 ITRF2008 Reference Network (Altamimi, Collilieux and Mtivier 2011)

ITRF coordinates are published as Cartesian equatorial X, Y and Z coordinates but can
be transformed to geographic coordinates (Latitude, Longitude and Height) referenced
to an ellipsoid. The IERS recommends the use of the Geodetic Reference System 1980
(GRS80 ellipsoid for this purpose (IGN 2015a). It is variations in these ellipsoidal
heights that this study will focus upon.

2.8 Vertical Reference Systems


Vertical datums are a means by which measured heights can be referenced against a
constant value or simply, a surface by which zero can be measured. Correctly assigning
a vertical datum can mean the difference between spatial data that is both accurate and
useable and spatial data that has little value.

23
Literature Review

2.8.1 The Geoid, Ellipsoid and Mean Sea Level


Traditionally national terrestrial datums tend be set at mean sea level (MSL), calculated
using one or more tide gauges over an extended period at a single location. It should be
acknowledged that even observing a full metonic cycle of 18.6 years would only
provide an approximation of true MSL as the dynamic influences of meteorological
conditions and ocean currents will remain. This disparity will likely have an impact on
the potential of the datum to accurately represent MSL.

The geoid can be defined as the surface of equipotential gravitational force that
corresponds to global mean sea level as considered independent of tidal and
environmental influence (OSi 2014). By definition gravitational force will act
perpendicular to the geoid and water bodies in equilibrium should rest parallel to the
geoid. The geoid is an irregular shape which can be approximated through mathematical
modelling and the use of harmonics (ESRI 2014). The Earths gravitational field is
affected by the Earths mass and the irregular distribution of the Earths mass results in
an irregular equipotential gravitational surface. Heights referenced to the geoid or to
MSL are orthometric heights. Orthometric heights refer to the perpendicular distance
between an object and the geoid.

An ellipsoid is a mathematical model that closely equates to the geoid, hence its use as a
GNSS reference figure. The geoid can deviate from the ellipsoid by approximately
100m in places globally (OSi 2014). This deviation is known as the Geoid-Ellipsoid
Separation (Figure 2.7).

Figure 2.7 Geoid-Ellipsoid Separation (http://principles.ou.edu/earth_figure_gravity/geoid/)

As previously stated (section 2.7) PPP solutions provide ellipsoidal heights referenced
to the GRS80 ellipsoid. Variations in these ellipsoidal heights will therefore be reflected
in any orthometric heights subsequently derived.

2.9 Python
Python is a community based programming language delivered through an open source
licence. Python contains a comprehensive object library and access to thousands of third
party modules (Python Software Foundation 2015a). Python is considered an
interpretive language as an interpreter is utilised to execute scripts and programs

24
Literature Review

(Downey 2012). Python is currently available in two forms, Python 2.x and Python 3.x
(Python Software Foundation 2015a). Although Python 3.x is the more modern, and
generally considered an improved language, it is not backward compatible and so many
modules developed through Python 2.x are not available in Python 3.x (Downey 2012).
This has ensured that migration of users from Python 2.x to Python 3.x has been
restricted. Python is a high level, object oriented programming language. Python is a
preferential programming language for the development of short scripts for data
manipulation and use in computational science due to the modularised nature and clean
syntax associated with the language (Langtangen 2008).

25
Methodology

3 Methodology
This chapter provides details of the methodology used to collect and process the data
required for the study. It includes details of the site selection process, data acquisition,
baseline establishment, processing of tidal data and GNSS processing using CSRS-PPP
and RTKLIB. The methodology will also discuss the various data management
techniques, focusing in particular on the use of Python scripts.

The study utilises GNSS and tidal data collected between 1st January and 14th January
2014. This period was selected as extreme tides were recorded (National Oceanography
Centre 2014), and the impact of OTL was likely to be enhanced. All data utilised in the
project came from established CORS GNSS stations and tide gauges. The project did
not require any field observations and focused entirely on the processing of existing
data.

The GNSS data collected for the selected sites was processed through RTKLIB and the
CSRS-PPP online service using three hour time-windows. The tidal data for each
location was extracted for a similar time-window. Different OTL modelling parameters
were applied to the PPP algorithms for each data set. The ellipsoidal heights for each
solution were then extracted using Python scripting. Baseline elevation for each site was
established using 24 hours of the same observation files processed through the
TrimbleRTX online service. The elevations generated by the different algorithm settings
were then statistically compared and analysed against each other, the baseline elevation,
and the corresponding tidal data, using Microsoft Excel.

Processing of RTKLIB solutions provided a significantly less time consuming process


than the processing of CSRS-PPP solutions. For this reason the use of the CSRS-PPP
service was limited to two sites. It was anticipated that comparison of the CSRS-PPP
solutions with the RTKLIB solutions for the corresponding sites would enable the likely
impact of OTL on CSRS-PPP solutions for the remaining sites to be estimated.

Although statistical data associated with the solutions was available, it was considered
beyond the scope of this project to examine such data. Similarly, the horizontal
component of each solution was not utilised in this study.

The author had no existing knowledge of the RTKLIB package prior to commencing the
study. Therefore, a significant amount of the project time was devoted to developing an
understanding of the application of PPP through the RTKLIB package, and to testing
the use of the various modelling parameters available within the package. This stage of
the project involved a highly iterative approach which tested various parameters and
settings. This element was not documented as the methodology utilised was ad hoc and
unscientific.

A simplified schematic of the methodology utilised for the project is illustrated in


Figure 3.1.

26
Methodology

Figure 3.8 Schematic Detailing the Project Methodology

3.1 Coordinate Reference System


All data was processed using PPP solutions and the results of these solutions were
available in ITRF at the epoch for which the data was collected. Therefore, there was no
requirement to undertake coordinate transformations during the project. This study has
focused entirely on the use of ellipsoidal heights derived from the ITRF and, therefore,
the ellipsoidal heights generated are based on the GRS80 Ellipsoid. This ensured that
the loss of accuracy that would normally be associated with coordinate system
transformations was minimised, and ensured the integrity of comparisons between the
processing methods.

3.2 Site Selection & Data Acquisition


The study focused upon coastal sites around Britain and Ireland. Sites were selected to
provide a broad distribution that would facilitate an assessment of the impact of OTL
around these coastal regions. Sites were also selected depending on the availability of
historical GNSS data over the study period. Proximity and availability of tide gauge
data for each site was also a considering factor.

27
Methodology

Three primary data sources were utilised in the project:

British Isles Continuous GNSS Facility

The UK Tide Gauge Network

The Irish National Tide Gauge Network

3.2.1 British Isles Continuous GNSS Facility


The British Isles Continuous GNSS Facility (BIGF) program is delivered by the
Nottingham Geospatial Institute of Nottingham University and funded by the Natural
Environment Research Council (NERC). The project commenced in 1998 and NERC
funding began in 2002. BIGF acts as a repository for CORS data generated by the
Ordnance Survey of Great Britain (OSGB), Ordnance Survey of Ireland (OSI), and
Land and Property Service of Northern Ireland (LPSNI). The service also provides
scientific researchers with access to the data. In addition to the CORS data from the
national mapping agencies, the BIGF program also stores GNSS data associated with
the UK Tide Gauge Networks tide gauge sites (University of Nottingham 2015). Data
requests are handled through an online request form
(http://www.bigf.ac.uk/request_data) and the files forwarded to the researcher via an
emailed download link. All GNSS observation files utilised in this project were sourced
from the BIGF program.

3.2.2 The UK Tide Gauge Network


The UK Tide Gauge Network (UKTGN) is a section of the UK Coastal Monitoring and
Forecasting (UKCMF) service. It consists of 43 tide gauges around the coasts of the
UK. The tide gauges are owned by the UK Environment Agency and maintained by the
National Oceanography Centre. Access to data from the network is provided by the
British Oceanographic Data Centre (BODC). BODC provide an online download
service for registered users in which access to all processed data is made available
(British Oceanographic Data Centre 2014).

All UK tidal data utilised in this project was retrieved through the BODC online
download system (https://www.bodc.ac.uk/data/online_delivery/ntslf/processed).

3.2.3 The Irish National Tide Gauge Network


The Irish National Tide Gauge Network (INTGN) is a collaborative project between
Irelands Marine Institute and various public and private organisations (Marine Institute
2015). Access to data from 14 tide gauges around the Irish coast is available through the
Marine Institute Data Portal (http://data.marine.ie/). The portal is free to use and
requires no user registration. The selected data is downloaded directly from the portal.

3.2.4 Sites
Six sites were selected for the study. These were located at:

St. Marys (Isles of Scilly)

Cork (Ireland)

28
Methodology

Belfast (Ireland)

Aberdeen (Scotland)

Lerwick (Shetland Islands)

North Shields (England)

The approximate location of each site is illustrated in Figure 3.2.

Figure 3.9 Approximate Location of Study Sites (Source: Google Earth )

The Lerwick, Aberdeen and North Shields sites all utilised GNSS and tide gauge data
collected from the UKTGN tide gauges. The Cork, Belfast and St Marys sites all
utilised data from tide gauges and national mapping agency CORS stations at different
locations. The St. Marys site should see no significant delay between the tides at the
tide gage and the location of the CORS station. The Belfast site utilised tides collected
at Bangor, Co. Down, approximately 15 km from the CORS station so no significant
tidal delay was envisaged.

The Cork site was more problematic as the nearest available tidal data was located at
Castletownbere, Co. Cork, approximately 100km from the CORS station at Cork.
However, the tidal delay between the two sites was estimated using various online tidal
charts to approximate a 20 to 30 minutes delay between high tide at Castletownbere and
high tide at Cork. This was deemed acceptable, considering the aims of the project, and
the three hour resolution at which the tidal data was to be considered.

29
Methodology

Details of the sites and the data sources for each are contained in Table 3.1.

Table 3.6 Data Details for Selected Sites

Site Code Location GNSS Data Tidal Data Comments


GNSS / Tide Gauge Source Source
St. Marys Minimal Tide
SCIL OSGB CORS UKTGN
(Scilly Isles) Delay
5-10 Minute
BELF Belfast / Bangor LPSNI CORS UKTGN
Tide Delay
Lerwick No Tide
LWTG UKTGN UKTGN
(Shetland Isles) Delay
No Tide
ABER Aberdeen UKTGN UKTGN
Delay
No Tide
NSLG North Shields UKTGN UKTGN
Delay
Cork / 20-30 Minute
CORK OSi CORS INTGN
Castletownbere Tide Delay

The site codes contained in Table 3.1 will be used for reference to the site through the
remainder of the text.

3.2.5 Tidal Data


The tidal date collected from both the BODC and Marine Institute Data Portal was
returned in text format. The resolution of the data was at 15 minute intervals.

3.2.6 GNSS Data


The GNSS data was provided in RINEX 2 format. It consisted of 24 hour RINEX files
for each site over the selected time period. An accompanying RINEX 2 navigation file
was also received for each date requested.

3.3 Baseline Elevation


Baseline elevations were established using 24 hours of GNSS data collected on
01/01/2014 for each site. These daily observation files were submitted to the
TrimbleRTX-PP online service (see section 2.4.2). Ellipsoidal heights for epoch 2014
were then manually extracted from the emailed results. Details of the results of the
TrimbleRTX-PP solutions are contained in Table

Insert Table Here showing positions

It acknowledged that the accuracy of the baseline ellipsoidal height would be


significantly improved had a significantly larger number of 24 hour windows been
processed and the mean value determined for each site. However, given the significant
time associated with manual uploads to the TrimbleRTX-PP service, and the inability to
automate the data extraction process, this was not undertaken.

30
Methodology

3.4 OTL Files Generation


The OTL files used in the project were generated using the Onsala Space Observatory
FLP (see section 2.6). This processed required the uploading of the station names and
locations in a precise format. The station names were provided to the service in the
same format as they appeared on the RINEX observation files associated with each site.
This ensured that the RTKLIB PPP solution would read the values into the PPP
algorithm correctly. The x, y, z coordinates for each site were extracted from the
TrimbleRTX-PP baselines solutions. The computed OTL files were then returned as
text within an email. This text was saved in a text document with .BLQ file extension.
BLQ is the standard OTL format utilised by the Onsala Space Observatory FLP
(Scherneck and Bos 2015).

Table 3.2 contains details of the different OTL models generated and the names given to
each model for the purposes of this study.
Table 3.7 OTL Models Generated for Study

Model File Tidal Model Correction for CMM


Name
GOT00.2_Solid GOT00.2 Not Applied
GOT00.2_CMM GOT00.2 Applied
GOT4.7 GOT4.7 Not Applied

3.5 Time-Window for GNSS Processing


It was decided that a three hour observation time period would be utilised for all CSRS-
PPP and RTLIB GNSSS processing. This time period was selected to ensure the
convergence time for the PPP solutions was adequate to provide solutions that were less
likely to absorb the effects of OTL within the noise elements of the solutions. It was
also deemed that the 3 hour time-window was sufficiently short to prevent OTL
displacements from being excessively averaged out, as would be expected if the
processed time-windows were excessively large.

3.6 CSRS PPP Processing


The CSRS-PPP online service (see section 2.4.1) was utilised for the processing of
seven days of observation files (from 01/01/2014 to 07/01/2014 inclusive) for the St.
Marys and Belfast sites only. CSRS-PPP processing was limited to these two sites only
as the magnitude of OTL was expected to vary significantly between the chosen sites.
This would allow analysis of the impact of OTL on CSRS-PPP solutions which could
be extrapolated to the other sites by comparison to the RTKLIB results.

Registration was required prior to use of the service. The data was uploaded to the sites
with varying OTL settings applied. Prior to upload the RINEX observation data for each
site was time-windowed to the required three hour period. This process utilised the
TEQC software application

31
Methodology

3.6.1 TEQC
The TEQC software program, which stands for Translate, Edit and Quality Check, is a
GNSS pre-processing tool developed by UNAVCO. TEQC provides a range of
functions including time-windowing, satellite filtering, format translations and quality
checks (UNAVCO 2015). TEQC uses a command line interface however a Windows
GUI called winTEQC Editor has been developed by Mark W. Huber of the US Army
Corps of Engineers Army Geospatial Centre. This GUI was utilised in this project and
was available for download from (http://teqc.silkwerks.com/). Figure 3.3 illustrates the
winTEQC GUI with a RINEX file ready for time-windowing to three hour windows.

Figure 3.10 winTEQC GUI

Prior to upload to the CSRS-PPP site, all the required daily observation files were
processed to three hour time-windows using the winTEQC Editor. These files were then
compressed to .zip format with each file containing the eight time-windowed RINEX
files for the corresponding day. These compressed files were then uploaded to the
CSRS-PPP service.

3.6.2 Ocean Tide Loading Parameters


The application of OTL parameters applied varied during the upload process through
use of OTL models generated for the study (see section 3.4). Details of the OTL models
applied are contained in Table 3.3.
Table 3.8 Details of CSRS-PPP Processing OTL Parameters

Site Name Dates OTL Model File Name


SCIL 01-07/01/2014 -
SCIL 01-07/01/2014 GOT00.2_Solid
SCIL 01-07/01/2014 GOT00.2_CMM
BELF 01-07/01/2014 GOT4.7
BELF 01-07/01/2014 -

32
Methodology

3.6.3 Data Extraction


The results of the CSRS-PPP processing for each 24 hour period were accessed via a
download link. The results were provided in two tiers of compressed folders in .zip
format as illustrated in Figure 3.4.

Figure 3.11 CSRS-PPP Results File Structure

To automate the extraction process a Python script was developed that would extract the
results files from the compressed formats, read the required ellipsoidal heights from
the .pos files for each solution and write these results to a text file that could be opened
and manipulated in Microsoft Excel.

Figure 3.5 illustrates the Python script as applied for the extraction of the BELF
GOT4.7 data.

Figure 3.12 Python Script for Extraction of BELF GOT4.7 CSRS-PPP Ellipsoidal Heights

The Python script utilised the following Python Modules:

Zipfile

Glob

OS

These modules were all available from the Python Standard Library (Python Software
Foundation 2015c). The script was modified for each site by editing the path directories
within the script. This provided an efficient and effective methodology for the
extraction of large volumes of data.

33
Methodology

The script was designed to extract only ellipsoidal heights but could easily be modified
to extract any data required. It was a pre-requirement of the script that the results
associated with each station and OTL file combination for the entire observation period
were stored in a single folder. Development and successful implementation of this
script ensured that Goal 1 of the thesis was achieved. A full copy of the script is
contained in Appendix A.

3.7 RTKLIB Processing


RTKLIB processing was undertaken for all sites included in the study. Seven days of
data from 01-01-2014 to 07-01-2014 were processed with and without OTL files
applied. The exception to this was SCIL for which 14 days of data, 01-01-2014 to 14-
01-2014, was processed. A single extended period was decided upon to provide a more
robust statistical analysis for one site. This would help identify if there were significant
statistical flaws in the use of the shorter time frames. Details of the tide models applied
for each site are contained in Table 3.4.

Table 3.9 Details of the OTL Modelling Files Applied During RTKLIB PPP Processing

Site Name Dates Number of Days OTL Model File Name


SCIL 01-14/01/2014 14 -
SCIL 01-14/01/2014 14 GOT00.2_Solid
SCIL 01-14/01/2014 14 GOT00.2_CMM
SCIL 01-14/01/2014 14 GOT4.7
BELF 01-07/01/2014 7 -
BELF 01-07/01/2014 7 GOT4.7
ABER 01-07/01/2014 7 -
ABER 01-07/01/2014 7 GOT4.7
LWTG 01-07/01/2014 7 -
LWTG 01-07/01/2014 7 GOT4.7
NSLG 01-07/01/2014 7 -
NSLG 01-07/01/2014 7 GOT4.7
CORK 01-07/01/2014 7 -
CORK 01-07/01/2014 7 GOT4.7

Application of PPP using RTKLIB required the collection of various correction and
modelling files (as discussed in section 2.3). It was decided that for consistency the use
of IGS products would take precedence over other available products. Differential Code
Bias (DCB) files from the Centre for Orbit Determination (CODE) were also utilised.
The products were accessed for the required time frame using the RTKGET application
(section 2.5.3) and applied in RTKPOST to generate the solutions.

Details of the products accessed through the RTKGET application and utilised in the
RTKPOST PPP solutions are contained in Table 3.5.

34
Methodology

Table 3.10 PPP Correction Files Utilised in RTKPOST PPP Solutions

File Correction Update


Product Type Provider
Extension Period
Satellite Clock Corrections IGS .clk Daily
Satellite Ephemeris Corrections IGS .sp3 Daily
Phase Centre Variations (PCV) IGS .atx Not Required
P1C1 Differential Code Biases
CODE .DCB Not Required
(DCB)

The RTKPOST PPP solution also required the application of a valid navigation file. The
daily navigation files that were associated with the GNSS observation files provided by
the BIGF were utilised for this purpose.

3.7.1 RTKPOST Settings


Once the positioning mode was set to PPP-static, most other parameters were left at the
default RTKPOST settings. The RTKPOST solutions were processed with GPS. This
was done as GLONASS navigation files could not be implemented concurrently with
the GPS navigation files. Figure 3.6 shows the RTKPOST settings used for the PPP
solutions with the OTL modelling facility not implemented.

35
Methodology

36
Methodology

Figure 3.13 RTKPOST PPP Solution Options Settings with OTL Off

The key changes to the RTKPOST settings included:

Processing Interval set to 3 hours.

Elevation Mask set to 10.

Earth Tides Corrections set to Solid (with OTL on/off as required).

Satellite/Receiver PCV and Phase Wind Up corrections were implemented.

Ionospheric Correction set to Iono-Free LC.

Tropospheric Correction set to Estimate ZTD + Grad.

Integer Ambiguity Resolution was Switched Off.

GDOP Threshold set to 20.

Solution for Static Mode set to Single.

37
Methodology

Ensuring all required IGS and CODE files were correctly implemented.

The use of EOP files within the solution was also investigated, however, it was found
that the results would produce significant outliers when ERP files were included. This
suggests that the ERP corrections have already been indirectly applied through the other
IGS products (clock and ephemeris corrections) as suggested by Kouba and Theroux
(2000). The inclusion of high-order ionospheric (TEC) files was found to have minimal
impact on the solutions and excluded from the study to facilitate a rapid processing
workflow was achieved. (CHECK THIS SENTENCE DOESN'T READ WELL)

The decision to implement a 10 elevation mask was based on examination of the


CSRS-PPP .sum files which indicated that a 10 elevation mask is implemented in the
CSRS-PPP algorithm. The application of the selected ionospheric and tropospheric
models was also supported by examination of the CSRS-PPP .sum files.

Selection of the GDOP value of 20 and the selection of the P1C1 .DCB files over the
other available options was based on ad hoc testing of the RTKPOST settings.

3.7.2 Data Extraction


The project required that the ellipsoidal heights were extracted from the various .pos
files generated as output from the RTKPOST processing. Extraction of the data was
automated through the use of a Python script. The script was developed through the use
of the Python Module Library (Python Software Foundation 2015c). The script
developed for the extraction of data from the RTKPOST solutions was significantly
simpler than that required for the CSRS-PPP results.

The script developed is illustrated in Figure 3.7.

Figure 3.14 Python Script for Extraction of RTKPOST Results

The script utilised the Python Glob module to create a list of all the files in a particular
directory with the .pos file extension. It then read the files in the list and extracted all
the solutions and wrote these to a single text file that was opened and manipulated in
Microsoft Excel. It was a requirement of the script that all the .pos results files
associated with a single site and PPP algorithm are stored in the same directory. This

38
Methodology

was (DELETE) requirement was automated in the RTKPOST process as the .pos results
files are automatically written to the folder containing the GNSS observation input files.
This script extracted all data associated with the PPP results file and not exclusively the
ellipsoidal heights. The ellipsoidal height data was, however, easily extracted from the
Excel workbook, as it was created as an individual column when converted to .csv
format.

Successful development and implementation of this Python script ensured that Project
Goal 2 was achieved.

3.8 Tidal Data Processing


The tidal data collected from the Marine Institute Data Portal and BODC was provided
in 15 minute intervals. This resolution was considered too high to facilitate successful
comparison and graphing in Microsoft Excel with the three hour resolution of the
processed GNSS positions. Therefore, a sub-sampling strategy was implemented to
reduce the tidal data to a similar resolution as the processed PPP results.

To facilitate the required sub-sampling, Python scripts were developed to extract the
tidal data at a three hour resolution. As the data formats used by the UKTGN and
INTGN tidal data were not consistent, two separate scripts were developed. The scripts
used to sub-sample the INTGN and UKTGN data are illustrated in Figures 3.8 and 3.9
respectively.

Figure 3.15 Python Script for Sub-Sampling INTGN Tidal Data to 3 Hour Resolution

39
Methodology

Figure 3.16 Python Script for Sub-Sampling UKTGN Tidal Data to 3 Hour Resolution

Both scripts utilised the Itertools module from the Python Module Library (Python
Software Foundation 2015c). The script associated with the extraction of UKTGN data
represented a significant improvement on the initial script used for INTGN data sub-
sampling, as it did not require two files to be written during the sampling. The second
script also included additional programming lines associated with the removal of the
units value M appended to the tidal magnitude number in the UKTGN tidal files. This
ensured that the tidal values would be recognised as numbers and not strings when
imported to a Microsoft Excel worksheet for analysis.

3.8.1 Tidal Phase Shift


Tidal Phase shifts were applied to the different tidal data by simply shifting the time
period associated with the tidal data. As the data was recorded every three hours and the
a (DELETE) tidal cycle of high and low tides was completed approximately once every
12 hours, a 3 hour time shift would represent approximately a 90 phase shift.

It is acknowledged that this technique is extremely simplified and that use of the high
resolution tidal data in conjunction with the principal lunar semi-diurnal constituent of
approximately 12 hours and 25 minutes (Clarke and Penna 2010) would provide more
precise results. However, it was deemed that this level of precision was beyond the
limitations of this study.

3.9 Collation of Processing Results


Following successful implementation of the various data collection, PPP processing and
data extraction techniques described in this methodology, the results were collated into
a series of Microsoft Excel worksheets. A separate worksheet was created for each site.
Each worksheet contained the sub-sampled tidal data and the extracted ellipsoidal
40
Methodology

heights from the PPP solutions implemented for that site. This would facilitate a
detailed statistical analysis of the data. Details of this analysis and presentation of the
results will be described in the next chapter.

A summary of the data generated from the methodology using the RTKPOST and
CSRS-PPP solutions is contained in Tables 3.6 and 3.7 respectively. Tidal data was
available for all sites for the entire study period.
Table 3.11 Summary of RTKPOST Processed Data for Each Study Site and the OTL Model Applied

SITE RTKPOST
CODE GOT00.2_Solid GOT00.2_CMM GOT4.7
SCIL 14 days 14 days 14 days
BELF - - 7 Days
LWTG - - 7 Days
ABER - - 7 Days
NSLG - - 7 Days
CORK - - 7 Days

Table 3.12 Summary of CSRS-PPP Processed Data for Each Study Site and the OTL Model Applied

SITE CSRS-PPP
CODE GOT00.2_Solid GOT00.2_CMM GOT4.7
SCIL 7days 7 days -
BELF - - 7 Days

41
Analysis and Results

4 Analysis and Results


This chapter details the statistical analysis of the data produced during the processes
described in the preceding chapter. All analyses will utilise observation values
normalised to the baseline height or mean tidal magnitude for each site. The results will
be illustrated using a combination of graphs and tables generated through Microsoft
Excel.

The first element of the analysis will attempt to establish the relationship between the
CSRS-PPP solutions and the RTKPOST solutions. This will be undertaken through
analysis and comparison of the results from the various implementations of the two
techniques at the SCIL and BELF sites.

Following this, the impact of the application of correcting for CMM in the OTL files
will be assessed. This will utilise the 14 days of RTKPOST data collected at SCIL using
the GOT00.2_Solid and GOT00.2_CMM OTL files. This analysis will also utilise the 7
days of data from SCIL processed through the CSRS-PPP service using the
GOT00.2_Solid and GOT00.2_CMM OTL files.

The next stage will involve an analysis of the impact of OTL at each site. This analysis
will focus on the use of RTKPOST solutions generated for all sites. It will examine
correlations between the various PPP solutions and the tidal data, including an analysis
of the significance of applying OTL files.

The final stage will attempt to identify if there is a significant phase lag between tides
and OTL at each site. This will entail an examination of the correlations between the
phase shifted tidal data and the RTKPOST solutions with no OTL applied.

A range of statistical tests will be utilised throughout the various analyses including:

Assessment of the Observation Range

Pearsons Correlation

F-Test: Two-Sample for Variances

t-Test: Paired Two-Sample for Means

The F-Test: Two-Sample for Variances tests were undertaken to compare the variances
of each pair of datasets. The hypotheses, tested in this study at a 95% confidence level,
are given as:

(Eq. 0)
Where:
= Variance of Dataset A
= Variance of Dataset B

42
Analysis and Results

A t-Test: Paired Two-Sample for Means will test, at a 95% confidence level, the null
hypotheses that the means of the two datasets can be considered equal.

4.1 CSRS-PPP versus RTKPOST Solutions


This section will examine the relationship between the various CSRS-PPP and
RTKPOST PPP solutions at the SCIL and BELF sites. It will compare the results of the
two techniques when applied using similar OTL parameters. Each pairing will be
graphed and the Pearsons Correlation coefficients calculated. The paired datasets will
then be analysed using an F-Test: Two-Sample for Variances and a t-Test: Paired Two-
Sample for Means.

A summary of the paired datasets to be analysed is contained in Table 4.1.


Table 4.13 Paired Datasets for CSRS-PPP v RTKPOST Analysis

SITE PAIRED DATASETS


SCIL RTKPOST No OTL CSRS-
PPP No
OTL
SCIL RTKPOST GOT00.2_Solid CSRS-
PPP
GOT00.2
_Solid
SCIL RTKPOST GOT00.2_CMM CSRS-
PPP
GOT00.2
_CMM
BELF RTKPOST GOT4.7 CSRS-
PPP
GOT4.7
BELF RTKPOST No OTL CSRS-
PPP No
OTL

4.1.1 SCIL: RTKPOST No OTL v CSRS-PPP No OTL


Figure 4.1 presents a 7 day time-series of the RTKPOST No OTL v CSRS-PPP No
OTL solutions at site SCIL from 01/01/2014 and 07/01/2014 inclusive. A total of 56
observations for each solution are included.

43
Analysis and Results

Figure 4.17 7 Day Time-Series Showing Normalised RTKPOST v CSRS-PPP No OTL Solutions at SCIL

Table 4.2 shows the Observation Range for each technique.


Table 4.14 Observation Range for RTKPOST and CSRS-PPP No OTL Solutions at SCIL

RTKPOST No OTL CSRS-PPP No OTL


Observation Range (m) 0.201 0.184

The range difference between the two processing methods was 0.017m. The CSRS-PPP
solutions have a smaller range; yet considering the number of observations in the
comparison and the potential for systematic noise in both solutions the observation
ranges can be considered reasonably comparable with respect to the accuracy of the
solutions.

The Pearsons Correlation between the two datasets was calculated at 0.88. This
suggests that there is a very strong relationship between the two datasets. This strong
correlation is clearly identifiable in Figure 4.1.

Results of the F-test analysis are contained in Table 4.3.


Table 4.15 F-Test: Two-Sample for Variances RTKPOST v CSRS-PPP No OTL Solutions at SCIL

RTKPOST No OTL CSRS-PPP No OTL


Mean -0.0133 -0.0019
Variance 0.0031 0.0025
Observations 56 56
df 55 55
F 1.225
P(F<=f) one-tail 0.227
F Critical one-tail 1.564

44
Analysis and Results

As the calculated F value of 1.225 is less than the critical value of 1.564 we can accept

H . Therefore the variances of the two datasets can be considered equal at a 95%
o

certainty level.

The results of the t-Test are contained in Table 4.4.


Table 4.16 t-Test: Paired Two-Sample for Means RTKPOST v CSRS-PPP No OTL Solutions at SCIL

RTKPOST No OTL CSRS-PPP No OTL


Mean -0.0133 -0.0019
Variance 0.0031 0.0025
Observations 56 56
Pearson Correlation 0.88
Hypothesized Mean 0
Difference
Df 55
t Stat -3.192
P(T<=t) one-tail 0.001
t Critical one-tail 1.673
P(T<=t) two-tail 0.002
t Critical two-tail 2.004
As the calculated t-Stat of -3.192 is outside the t Critical two-tail range of +/- 2.004 the
two means cannot be considered equal and the null hypotheses must be rejected.

4.1.2 SCIL: RTKPOST GOT00.2_Solid v CSRS-PPP GOT00.2_Solid


Figure 4.2 presents a 7 day time-series of the RTKPOST GOT00.2_Solid v CSRS-PPP
GOT00.2_Solid solutions at site SCIL from 01/01/2014 and 07/01/2014 inclusive. A
total of 56 observations for each solution are included.

45
Analysis and Results

Figure 4.18 7 Day Time-Series Showing Normalised RTKPOST v CSRS-PPP GOT00.2_Solid Solutions at
SCIL

Table 4.5 shows the Observation Range calculated for each technique.
Table 4.17 Observation Range for RTKPOST and CSRS-PPP GOT00.2_Solid Solutions at SCIL

RTKPOST CSRS-PPP
GOT00.2_Solid GOT00.2_Solid
Observation Range (m) 0.112 0.052

There is a 0.06m difference between the two observation ranges. The CSRS-PPP
solutions demonstrate a significantly greater degree of accuracy than the RTKPOST
solutions. The validity of this conclusion is supported by visual examination of the
time-series plot (Figure 4.2) from which no significant outliers can be identified.

The Pearsons Correlation between the two datasets was calculated at -0.3. This
suggests that there is little correlation between the two datasets.

Results of the F-test analysis are contained in in Table 4.6.

Table 4.18 F-Test: Two-Sample for Variances RTKPOST v CSRS-PPP GOT00.2_Solid Solutions at SCIL

RTKPOST GOT00.2_Solid CSRS-PPP GOT00.2_Solid


Mean -0.0109 -0.0014
Variance 0.0008 0.0001
Observations 56 56
Df 55 55
F 6.610
46
Analysis and Results

P(F<=f) one-tail 0.000


F Critical one-
1.564
tail

As the calculated F value of 6.61 is significantly higher than the critical value of 1.564

there is very strong reasons to reject H . Therefore the variances of the two datasets
o

cannot be considered equal.

The results of the t-Test are contained in Table 4.7.


Table 4.19 t-Test: Paired Two-Sample for Means RTKPOST v CSRS-PPP GOT00.2_Solid at SCIL

RTKPOST CSRS-PPP
GOT00.2_Solid GOT00.2_Solid
Mean -0.0109 -0.0014
Variance 0.0008 0.0001
Observations 56 56
Pearson Correlation -0.30
Hypothesized Mean Difference 0
Df 55
t Stat -2.188
P(T<=t) one-tail 0.016
t Critical one-tail 1.673
P(T<=t) two-tail 0.033
t Critical two-tail 2.004

As the calculated t-Stat of -2.188 is outside the t Critical two-tail range of +/- 2.004 the
two means cannot be considered equal and the null hypotheses must be rejected.

4.1.3 SCIL: RTKPOST GOT00.2_CMM v CSRS-PPP GOT00.2_CMM


Figure 4.3 presents a 7 day time-series of the RTKPOST GOT00.2_CMM v CSRS-PPP
GOT00.2_CMM solutions at site SCIL from 01/01/2014 and 07/01/2014 inclusive. A
total of 56 observations for each solution are included.

47
Analysis and Results

Figure 4.19 7 Day Time-Series Showing Normalised RTKPOST v CSRS-PPP GOT00.2_CMM Solutions at
SCIL

Table 4.8 shows the Observation Range calculated for each technique.
Table 4.20 Observation Range for RTKPOST and CSRS-PPP GOT00.2_CMM Solutions at SCIL

RTKPOST CSRS-PPP
GOT00.2_CMM GOT00.2_CMM
Observation Range (m) 0.109 0.048

There is a 0.059m difference between the two observation ranges. The CSRS-PPP
solutions demonstrate a significantly greater degree of accuracy than the RTKPOST
solutions. The validity of this conclusion is supported by visual examination of the
time-series plot (Figure 4.3) from which no significant outliers can be identified in
either dataset.

The Pearsons Correlation between the two datasets was calculated at -0.09. This
suggests that there is no correlation between the two datasets.

Results of the F-test analysis are contained in Table 4.9.


Table 4.21 F-Test: Two-Sample for Variances RTKPOST v CSRS-PPP GOT00.2_CMM Solutions at SCIL

RTKPOST GOT00.2_CMM CSRS-PPP GOT00.2_CMM


Mean -0.0139 -0.0012
Variance 0.0007 0.0001
Observations 56 56
Df 55 55
F 6.188
P(F<=f) one-tail 0.000
F Critical one- 1.564
tail

48
Analysis and Results

As the calculated F value of 6.188 is significantly higher than the critical value of 1.564

there is very strong reason to reject H . Therefore the variances of the two datasets
o

cannot be considered equal.

The results of the t-Test are contained in Table 4.10.


Table 4.22 t-Test: Paired Two-Sample for Means RTKPOST v CSRS-PPP GOT00.2_CMM at SCIL

RTKPOST CSRS-PPP
GOT00.2_CMM GOT00.2_CMM
Mean -0.014 -0.001
Variance 0.001 0.000
Observations 56 56
Pearson Correlation -0.092
Hypothesized Mean Difference 0
Df 55
t Stat -3.267
P(T<=t) one-tail 0.001
t Critical one-tail 1.673
P(T<=t) two-tail 0.002
t Critical two-tail 2.004

As the calculated t-Stat of -3.267 is outside the t Critical two-tail range of +/- 2.004 the
two means cannot be considered equal and the null hypotheses must be rejected.

4.1.4 BELF: RTKPOST No OTL v CSRS-PPP No OTL


Figure 4.4 presents a 7 day time-series of the RTKPOST No OTL v CSRS-PPP No
OTL solutions at site BELF from 01/01/2014 and 07/01/2014 inclusive. A total of 56
observations for each solution are included.

49
Analysis and Results

Figure 4.20 7 Day Time-Series Showing Normalised RTKPOST v CSRS-PPP No OTL Solutions at BELF

Table 4.11 shows the Observation Range calculated for each technique.
Table 4.23 Observation Range for RTKPOST and CSRS-PPP No OTL Solutions at BELF

RTKPOST CSRS-PPP
GOT00.2_CMM GOT00.2_CMM
Observation Range (m) 0.134 0.080

There is a 0.054m difference between the two observation ranges. The CSRS-PPP
solutions demonstrate a significantly greater degree of accuracy than the RTKPOST
solutions. The validity of this conclusion is supported by visual examination of the
time-series plot (Figure 4.4) from which no significant outliers can be identified in
either dataset.

The Pearsons Correlation between the two datasets was calculated at 0.66. This
suggests that there is a moderate correlation between the two datasets. Although the
statistical value is inconclusive, the argument that there is a relationship between the
datasets is supported by visual examination of Figure 4.4. It is possible that the
RTKPOST elevation at 06:00AM on 03/01/14 (circled in Figure 4.4), which is 0.069m
above the baseline elevation, may have detracted significantly from the statistical
correlation. As it is poor scientific practice to manipulate the data without just cause, no
further action was taken.

Results of the F-test analysis are contained in Table 4.12.

50
Analysis and Results

Table 4.24 F-Test: Two-Sample for Variances RTKPOST v CSRS-PPP No OTL Solutions at BELF

RTKPOST No OTL CSRS-PPP No OTL


Mean -0.0002 -0.0007
Variance 0.0006 0.0003
Observations 56 56
df 55 55
F 1.804
P(F<=f) one-tail 0.015
F Critical one-tail 1.564

As the calculated F value of 1.804 is higher than the critical value of 1.564 there is good
reason to reject Ho. Therefore, the variances of the two datasets cannot be considered
equal.

The results of the t-Test are contained in Table 4.13.


Table 4.25 t-Test: Paired Two-Sample for Means RTKPOST v CSRS-PPP No OTL Solutions

RTKPOST No OTL CSRS-PPP No OTL


Mean -0.0002 -0.0007
Variance 0.0006 0.0003
Observations 56 56
Pearson Correlation 0.661
Hypothesized Mean 0.000
Difference
Df 55
t Stat 0.180
P(T<=t) one-tail 0.429
t Critical one-tail 1.673
P(T<=t) two-tail 0.858
t Critical two-tail 2.004

As the calculated t-Stat of 0.18 is well within the t Critical two-tail range of +/- 2.004,
the two means can be considered statistically equal and the null hypotheses can be
accepted.

4.1.5 BELF: RTKPOST GOT4.7 v CSRS-PPP GOT4.7


Figure 4.5 presents a 7 day time-series of the RTKPOST GOT4.7 v CSRS-PPP GOT4.7
solutions at site BELF from 01/01/2014 and 07/01/2014 inclusive. A total of 56
observations for each solution are included.

51
Analysis and Results

Figure 4.21 7 Day Time-Series Showing Normalised RTKPOST v CSRS-PPP GOT4.7 Solutions at BELF

Table 4.14 shows the Observation Range calculated for each technique.
Table 4.26 Observation Range for RTKPOST and CSRS-PPP GOT4.7 Solutions at BELF

RTKPOST GOT4.7 CSRS-PPP GOT4.7


Observation Range (m) 0.141 0.056

There is a 0.085m difference between the two observation ranges. The CSRS-PPP
solutions demonstrate a significantly greater degree of accuracy than the RTKPOST
solutions. Several significant deviations do exist in the RTKPOST solution (see Figure
4.5); however, there is no sound scientific reason to exclude these observations from the
analysis.

The Pearsons Correlation between the two datasets was calculated at 0.47. This figure
suggests that there is a little correlation between the two datasets.

Results of the F-test analysis are contained in Table 4.15.


Table 4.27 F-Test: Two-Sample for Variances RTKPOST v CSRS-PPP No OTL Solutions at BELF

RTKPOST GOT4.7 CSRS-PPP GOT4.7


Mean -0.0041 -0.0004
Variance 0.0005 0.0001
Observations 56 56
Df 55 55
F 3.235
P(F<=f) one-tail 0.0000
F Critical one- 1.564
tail

52
Analysis and Results

As the calculated F value of 3.235 is significantly higher than the critical value of 1.564
there is very strong reason to reject Ho. Therefore, the variances of the two datasets
cannot be considered equal.

The results of the t-Test are contained in Table 4.16.


Table 4.28 t-Test: Paired Two-Sample for Means RTKPOST v CSRS-PPP No OTL Solutions

RTKPOST No OTL CSRS-PPP No OTL


Mean -0.0041 -0.0004
Variance 0.0005 0.0001
Observations 56 56
Pearson Correlation 0.48
Hypothesized Mean 0
Difference
Df 55
t Stat -1.449
P(T<=t) one-tail 0.0766
t Critical one-tail 1.6730
P(T<=t) two-tail 0.1531
t Critical two-tail 2.004
As the calculated t-Stat of -1.449 is within the t Critical two-tail range of +/- 2.004 the
two means can be considered statistically equal and the null hypotheses can be accepted.

4.1.6 Summary of Findings


A summary of the findings of the various statistical tests between the CSRS-PPP and
RTKPOST solutions from the SCIL and BELF sites are presented in Table 4.12.
Table 4.29 Summary of Statistical Comparison of RTKPOST v CSRS-PPP Solutions

Range Difference
Correlation F-Test for t-Test for
Site OTL Model RTKPOST minus
Coefficient Variance Means
CSRS-PPP
SCIL None 0.017m 0.88
SCIL GOT00.2_Solid 0.060m -0.3
SCIL GOT00.2_CMM 0.059m -0.09
BEL None 0.054m
0.66
F
BEL GOT4.7 0.085m
0.48
F

Examination of the normalised range differences between the CSRS-PPP and


RTKPOST solutions indicates that the CSRS-PPP solutions have a significantly greater
accuracy level than the RTKPOST solutions. The exception to this are the results from
the PPP processing at SCIL with no OTL models applied. Although the accuracy of
these CSRS-PPP solutions is better than the RTK-POST solutions, the difference is
much less pronounced. The anticipated OTL at SCIL is greater than that at BELF

53
Analysis and Results

(Clarke and Penna 2010) and the large normalised ranges associated with the two
techniques at SCIL when OTL modelling is not applied support this assumption. It can
be suggested from these results that the magnitude of the OTL displacement at SCIL
absorbs the remaining systematic noise differences between the two techniques when
OTL is not modelled.

The correlation values suggest that there is a reasonably good correlation between the
RTKPOST and CSRS-PPP solutions at both sites when utilised without applying OTL
models. The correlation between the two techniques when OTL models are applied falls
significantly. This suggests that the application of OTL models systematically removes
the causing factor for the correlation between the two techniques. As the OTL models
are designed to mitigate OTL displacements, it is suggested that the underlying cause
for the correlation is the OTL displacement.

The results of the F-Tests for Variance suggest that the variances of solutions using
CSRS-PPP and RTKPOST cannot be expected to be equal. The results have shown that
the lower variances associated with the CSRS-PPP results are statistically significant.
This indicates that the CSRS-PPP solutions have a greater precision that the RTKPOST
solutions.

The results of the t-Tests for Means are inconclusive. The two data pairs from the BELF
site show that the means can be statistically considered equal. However, the means from
all the SCIL datasets could not be considered equal. As the expected magnitude of OTL
at BELF is significantly lower than that at SCIL there is a potential to draw some
significance from these results; however, there is not adequate evidence to draw a
scientifically sound conclusion. This may be an area for future study utilising additional
sites and longer time-series data.

In conclusion, these results indicate that there is a significant inter-relationship between


the results of the use of RTKPOST and CSRS-PPP when considering solutions for
which OTL modelling has not been applied. This inter-relationship is strongest when
OTL displacement is of a greater magnitude.

54
Discussion

5 Discussion
5.1 Recommendations for Future Studies
Future studies should also consider the standard deviations of the results to
examine the impact of OTL.

A fully documented and scientific approach to ascertain the optimal RTKPOST


settings and modelling parameters should be undertaken.

55
Conclusion

6 Conclusion

56
References

References
Ahmed, T. 2014. RE: [RTX-PP Contact page] Thesis on effects of Ocean Tide Loading and PPP.
Type to QUINE, P.

Altamimi, Z., Collilieux, X. & Mtivier, L. 2011. ITRF2008: an improved solution of the
international terrestrial reference frame. Journal of Geodesy, 85, 457-473.

Bos, M. & Scherneck, H.-G. 2015. A short introduction to ocean tide loading provider [Online].
Sweden: Onsala Space Observatory. Available:
http://holt.oso.chalmers.se/loading/loadingprimer.html [Accessed February 2015].

British Oceanographic Data Centre. 2014. UK Tide Gauge Network at BODC [Online]. BODC.
Available: http://www.bodc.ac.uk/projects/uk/ntslf/.

Center for Operational Oceanographic Products and Services. 2013. Our Restless Tides - NOAA
Tides & Currents [Online]. USA: NOAA. Available: http://co-ops.nos.noaa.gov/restles5.html
[Accessed February 2015].

Clarke, P. J. & Penna, N. T. 2010. Ocean Tide Loading and Relative GNSS in the British Isles.
Survey Review, 42, 212-228.

Colombo, O. 1986. Ephemeris errors of GPS satellites. Bulletin godsique, 60, 64-84.

Doucet, K., Herwig, M., Kipka, A., Kreikenbohm, P., Landau, H., Leandro, R., Moessmer, M. &
Pagels, C. 2012. Introducing Ambiguity Resolution in Web Hosted Global Multi-GNSS Precise
Positioning with Trimble RTX-PP. Germany: Trimble TerraSat GmbH.

Downey, A. B. 2012. Think Python. US: O'Reilly Media.

Egbert, G. & Erofeeva, S. 2002. TPXO6.2 Load Tide Model.

El-Mowafy, A. 2012. Precise Real-Time Positioning using Network RTK. In: PROF. SHUANGGEN
JIN (ed.) Global Navigation Satellite Systems: Signal, Theory and Applications. InTech.

ESA. 2011a. DGNSS Fundamentals - Navipedia [Online]. Available:


http://www.navipedia.net/index.php/DGNSS_Fundamentals [Accessed February 2015].

ESA. 2011b. Differential GNSS - Navipedia [Online]. Available:


http://www.navipedia.net/index.php/Differential_GNSS [Accessed February 2015].

ESA. 2011c. GLONASS General Introduction - Navipedia [Online]. Available:


http://www.navipedia.net/index.php/GLONASS_General_Introduction [Accessed February
2015].

ESA. 2011d. GPS Future and Evolutions - Navipedia [Online]. Available:


http://www.navipedia.net/index.php/GPS_Future_and_Evolutions [Accessed February 2015].

ESA. 2011e. PPP Fundamentals - Navipedia [Online]. Available:


http://www.navipedia.net/index.php/PPP_Fundamentals [Accessed April 2015].

ESA. 2011h. Precise Point Positioning - Navipedia [Online]. Available:


http://www.navipedia.net/index.php/Precise_Point_Positioning [Accessed February 2015].

57
References

ESA. 2014. GPS Signal Plan - Navipedia [Online]. Available:


http://www.navipedia.net/index.php/GPS_Signal_Plan [Accessed February 2015].

ESRI. 2014. 1-Mean Sea Level, GPS, and the Geoid [Online]. Available:
http://www.esri.com/news/arcuser/0703/geoid1of3.html [Accessed 1st December 2014
2014].

Gao, Y. 2006. GNSS Solutions: Precise Point Positioning and its Challenges. Inside GNSS. USA:
Gibbons Media & Research LLC.

Ghilani, C. D. & Wolf, P. R. 2012. Elementary Surveying: An Introduction to Geomatics, US,


Prentice Hall.

Grinter, T. & Janssen, V. Post-Processed Precise Point Positioning: A Viable Alternative? 17th
Association of Public Authority Surveyors Conference (ASPAS 2012), 19-21 March, 2012
Wollongong, New South Wales, Australia.

Grinter, T. & Roberts, C. 2011. Precise Point Positioning: Where are we now? IGNSS
Symposium 2011. Sydney, Australia, 15-17 November.

IGN. 2015a. FAQ [Online]. Available: http://itrf.ensg.ign.fr/faq.php?type=answer [Accessed


April 2015].

IGN. 2015b. General concepts [Online]. Available: http://itrf.ensg.ign.fr/general.php [Accessed


April 2015].

IGS. 2015a. IGS About [Online]. Available: http://igs.org/about [Accessed February 2015].

IGS. 2015b. IGS Products [Online]. Available: http://igs.org/products [Accessed February 2015].

Jeffrey, C. 2010. An Introduction to GNSS: GPS, GLONASS, Galileo and other Global Navigation
Satellite Systems, Alberta, Canada, Novatel Inc.

Kouba, J. 2009. A Guide to Using International GNSS Service (IGS) Products. Canada: Natural
Resources Canada.

Kouba, J. & Theroux, P. 2000. GPS Precise Point Positioning using IGS Orbit Products. Ontario,
Canada: Nautural Resources Canada.

Langtangen, H. 2008. Python Scripting for Computational Science. Berlin: Springer-Verlag.

Leick, A. 2004. GPS Satellite Surveying, USA, John Wiley & Sons Inc.

Marine Institute. 2015. Tidal Observations | Marine Institute [Online]. Ireland. Available:
http://www.marine.ie/Home/site-area/data-services/real-time-observations/tidal-
observations [Accessed April 2015].

National Oceanography Centre. 2014. Storm surge hits UK Coastline | News | National
Oceanography Centre | from coast to deep ocean [Online]. Natural Environment Research
Council. Available: http://noc.ac.uk/news/storm-surge-hits-uk-coastline.

Natural Resources Canada. 2014. Tools and Applications | Natural Resources Canada [Online].
Available: http://www.nrcan.gc.ca/earth-sciences/geomatics/geodetic-reference-
systems/tools-applications/10925#ppp [Accessed February 2015].

58
References

O'Mahony, K. 2014. An Evaluation of Precise Point Positioning (PPP) and an Analysis of the
Accuracies Achieved by using Online PPP Post-Processing Services. BSc (Hons) Geomatics
(Surveying & Mapping), Dublin Institute of Technology.

OSi. 2014. Ordnance Survey Ireland | Geoid Model [Online]. Available: /Services/GPS-
Services/Reference-Information/Geoid-Model.aspx [Accessed December 2014 2014].

Python Software Foundation. 2015a. About Python [Online]. Available:


https://www.python.org/about/.

Python Software Foundation. 2015c. The Python Standard Library [Online]. Available:
https://docs.python.org/2/library/ [Accessed April 2015].

Ray, R. D. 1999. A Global Ocean Tide Model from TOPEX/Poseiden Altimetry: GOT99.2.
Available: http://ntrs.nasa.gov/search.jsp?R=19990089548 [Accessed April 2014].

Rizos, C., Janssen, V., Roberts, C. & Grinter, T. 2012. Precise Point Positioning: Is the Era of
Differential GNSS Positioning Drawing to an End? FIG Working Week 2012. Rome, Italy, 6-10
May.

Russian Institute of Space Device Engineering 2008. GLONASS Interface Control Document:
Navigational Radio Signal in Bands L1, L2. Moscow.

Sagatun, B. 2015. Positioning Systems Overview - Fugro Survey [Online]. Available:


http://www.starfix.com/positioning-systems/ [Accessed February 2015].

Scherneck, H.-G. 2011. CMC: Centre of Mass Correction [Online]. Gothenburg. Available:
http://holt.oso.chalmers.se/loading/cmc.html [Accessed April 2015.

Scherneck, H.-G. & Bos, M. S. 2015. Ocean tide loading provider [Online]. Available:
http://holt.oso.chalmers.se/loading/.

Scherneck, H. G. 1991. A parametrized solid earth tide model and ocean tide loading effects for
global geodetic baseline measurements. Geophysical Journal International, 106, 677-694.

Shum, K. P., Woodworth, P. L., Andersen, O. B., Egbert, G. D., Francis, O., King, C., Klosko, S. M.,
le Provost, C., Li, X., Molines, J.-M., Parke, M. E., Ray, R. D., Schlax, M. G., Stammer, D., Tierney,
C. C., Vincent, P. & Wunsch, C. I. 1997. Accuracy Assessment of Recent Ocean Tide Models.
Journal of Geophysical Research, 102, 25.

Silver, M. 2013. Seven Free Alternatives to OPUS GPS Post-Processing During U.S. Federal
Government Shutdown : GPS World [Online]. GPS World. Available: http://gpsworld.com/7-
free-alternatives-opus-post-processing-in-government-shutdown/ [Accessed April 2015].

Strang, G. & Borre, K. 1997. Linear Algebra, Geodesy, and GPS, U.S., Wellesley-Cambridge
Press.

Subirana, J. S., Zornoza, J. M. J. & Hernndez-Pajares, M. 2013. GNSS Data Processing,


Netherlands, ESA Communications.

Takasu, T. 2007. RTKLIB.

Takusa, T. 2013. RTKLIB ver. 2.4.2 Manual.

59
References

Throux, P. & Kouba, J. 2001. GPS precise point positioning using IGS orbit products. Physics
and Chemistry of the Earth, Part A: Solid Earth and Geodesy, 26, 573-578.

Trimble. 2015. Trimble CenterPoint RTX Post-Processing Service [Online]. Available:


http://www.trimblertx.com/ [Accessed April 2015].

UK Hydrographic Office. 2006. Harmonic Constant Specification. Available:


http://www.ukho.gov.uk/AdmiraltyPartners/FGHO/Pages/TidalHarmonics.aspx.

UNAVCO. 2015. TEQC [Online]. [Accessed April 2015].

University of Nottingham. 2015. bigf - The University of Nottingham [Online]. University of


Nottingham. Available: http://www.nottingham.ac.uk/ngi/research/engineering-surveying-
and-remote-measurement/projects/bigf.aspx.

Van De Marel, H. & De Baker, P. F. 2012. GNSS Solutions: Single- versus Dual-Frequency Precise
Point Positioning. Inside GNSS. USA: Gibbons Media & Research LLC.

Weston, N. D. & Schieger, V. 2010. Cost Effective GNSS Positioning Techniques. International
Federation of Surveyors (FIG).

60

You might also like