Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Case Study

Proposal for Geoid Model Evaluation from GNSS-INS/Leveling


Data: Case Study along a Railway Line in Greece
Vassilis Gikas1; Athanasios Mpimis2; and Aggeliki Androulaki3
Downloaded from ascelibrary.org by Universidad Nacional De Ingeniería on 12/28/20. Copyright ASCE. For personal use only; all rights reserved.

Abstract: A surveying-engineering procedure is presented for gravimetric geoid model verification along selected railway traverse lines using
collocated ellipsoid and orthometric heights. The proposed method relies on (1) a tactical-grade, multisensor [Global Navigation Satellite Sys-
tem (GNSS)/inertial navigation system (INS)] kinematic-surveying system used to measure the ellipsoid heights along the traveled section and
(2) the orthometric heights derived from the railway-line construction and/or maintenance works. Compared with the standard static GPS/
leveling technique, the proposed approach produces geoid height profiles in a very short acquisition time, with dense point spacing, and in
a considerably inexpensive manner. The quality of the final results reflects the accuracy of modern GNSS/INS systems and available rail-
track orthometric height information. The method was applied to a long (101-km) railway traverse in central Greece using a custom-built mobile
mapping system. Gravimetric geoid model verification was undertaken in an absolute and relative manner for the Earth Gravitational Model
2008 (EGM08) and 1996 (EGM96) geopotential models. Analysis of the results has demonstrated the potential of the method both in terms of
efficiency and in terms of the evaluation checks performed. Based on the absolute- and relative-quality tests undertaken, the superiority of
EGM08 over EGM96 is clearly evident for the tested line. Also, these results are cross compared and evaluated with the findings obtained from
previous studies in Greece. DOI: 10.1061/(ASCE)SU.1943-5428.0000099. © 2013 American Society of Civil Engineers.
CE Database subject headings: Height; Global positioning systems; Rail transportation; Surveys; Greece; Case studies.
Author keywords: Height; Geoid; GPS/leveling; GPS/INS; Spirit leveling; Railway survey; Greece.

Introduction systematic errors in satellite altimetry, and a bad elevation datum


used by gravity observations (Sideris et al. 1992; Fotiou et al. 2010;
A global geopotential model (GGM) constitutes a mathematical Rabah and Kaloop 2011; Erol 2012). Although the effects of these
representation of the Earth’s gravitational field in the form of a error sources have significantly decreased after the deployment of
spherical harmonic model defined up to a specific degree and order. the CHAMP, GRACE, and GOCE (National Aeronautics and Space
Its coefficients are obtained using a combination of source in- Administration 2004; GFZ German Research Centre for Geo-
formation including satellite orbit analyses, terrestrial gravity mea- sciences 2006; European Space Agency 2009) satellite gravity
surements, and satellite altimetry data over the oceans (Erol and Erol missions, the availability and geographic distribution of gravity
2012). With the advent of satellite positioning systems in the last two information worldwide plays a major role in the quality of GGM-
decades, the importance and practical use of GGMs has been ex- based heights obtained on different parts of the Earth. Therefore,
panded rapidly, as they are required in Global Positioning System validating GGMs with in situ data (such as gravity observations,
(GPS) surveys to convert ellipsoid heights to orthometric heights for gravity anomalies, GPS/leveling data, and astrogeodetic vertical
use in geodetic surveying and land-mapping applications. In this deflections) is a highly important task for obtaining an independent
process, the quality of the derived orthometric heights relies heavily estimate of their actual accuracy at different locations around the
on the quality of the geoid heights that are obtained as a function Earth and at various spatial scales.
of the geopotential coefficients that define a GGM. In fact, geoid In the past, numerous studies have been published that evaluate
heights may suffer several biases attributed to long-wavelength the accuracy of major GGMs in different regions of the Earth. This
errors contributed by geopotential model errors, uneven geo- work refers to evaluation studies undertaken by Amos and Feath-
graphic distribution and coverage of terrestrial gravity data, erstone (2003), Kiamehr and Sjöberg (2005), and Ustun and Demirel
(2006) in Australia, Iran, and Turkey, respectively. More recently,
1
Assistant Professor, School of Rural and Surveying Engineering, with the release of the combined Earth Gravitational Model 2008
National Technical Univ. of Athens, Athens 15780, Greece (corresponding (EGM08), most researchers have concentrated their efforts on per-
author). E-mail: vgikas@central.ntua.gr formance tests undertaken at a regional scale using various types of
2
Ph.D. Candidate, School of Rural and Surveying Engineering, external data, for example, studies by Erol et al. (2009), Sadiq and
National Technical Univ. of Athens, Athens 15780, Greece. E-mail: Ahmad (2009), Dawod et al. (2010), Ellmann (2010), and Martín
ampimis@central.ntua.gr et al. (2010) undertaken in Turkey, Pakistan, Egypt, the Baltic
3
M.S. Student, School of Rural and Surveying Engineering, National countries, and Spain, respectively.
Technical Univ. of Athens, Athens 15780, Greece. E-mail: aaggeliki@central. In Greece, a number of papers have been published in recent
ntua.gr
years that study the height systems in the region and attempt a
Note. This manuscript was submitted on August 10, 2012; approved on
November 16, 2012; published online on November 20, 2012. Discussion combined analysis of heterogeneous height data to obtain an im-
period open until October 1, 2013; separate discussions must be submitted proved local geoid that would lead to better vertical control
for individual papers. This paper is part of the Journal of Surveying (Delikaraoglou and Mintourakis 2010; Tziavos et al. 2010). Also,
Engineering, Vol. 139, No. 2, May 1, 2013. ©ASCE, ISSN 0733-9453/ several studies were undertaken that validate the performance
2013/2-95–104/$25.00. of GGMs and regional gravimetric geoid models in terms of

JOURNAL OF SURVEYING ENGINEERING © ASCE / MAY 2013 / 95

J. Surv. Eng., 2013, 139(2): 95-104


GPS/leveling observations (Andritsanos et al. 2000; Kotsakis et al. degree 2,190 and order 2,159. This model is a combined GGM; it is
2011; Tziavos et al. 2012). From these studies, the most extensive based on satellite gravity data (ITG-GRACE03S GRACE model),
work refers to a test network of 1,542 GPS/leveling benchmarks surface gravity anomaly data sampled over a 5 3 5 grid around
spanning mainland Greece (Kotsakis and Katsambalos 2010; Kotsakis the globe, digital elevation information (SRTM, GTOPO30, ICESat
et al. 2010). databases, etc.), and mean sea level data (Jason-1, ER1/2, Envisat
In this regard, the objective of this paper is twofold: firstly, to missions, etc.). However, it does not include any GPS/leveling data,
introduce an alternative to the standard surveying-engineering and thus, its evaluation based on similar information can provide an
(i.e., static GPS/leveling technique) approach for GGM testing independent assessment of its accuracy in different regions (Kenyon
and to assess its operational efficiency using real data; and secondly, et al. 2007; Pavlis et al. 2008, 2012).
to evaluate the performance of EGM08 and Earth Gravitational
Model 1996 (EGM96) along the surveyed section. Also, cross
comparisons of the results obtained with the findings from previous Methodological Approach
Downloaded from ascelibrary.org by Universidad Nacional De Ingeniería on 12/28/20. Copyright ASCE. For personal use only; all rights reserved.

studies in Greece are discussed. As opposed to the standard vali-


Traditionally, geoid heights (geoid-ellipsoid separation) are com-
dation procedure adopted in previous studies worldwide (i.e., single-
puted based on the gravimetric approach, also known as the
point comparisons between GGM- and GPS/leveling-derived
Stokes-Helmert method (Vanícek and Martinec 1994). Alterna-
undulations at benchmarks spanning a region of interest), this work
tively, geoid heights can be computed geometrically using the
introduces a kinematic data-acquisition procedure that allows the
following relation:
extraction of geoid separation along a lengthy railway section. In
our approach, as base information, we use the known orthometric N h2H ¼ h 2 H ð3Þ
heights from the rail-track construction and their ellipsoid height
equivalents obtained using a tactical-grade (military-grade), multi-
where h 5 ellipsoid height (above the ellipsoid); and H 5 Helmert-
sensor kinematic-surveying system. In this manner, we produce geoid
type orthometric height (above the geoid) (Fig. 1). In this com-
undulations for a dense and continuous traverse section, which allows
putation, the ellipsoid height is measured from GPS surveying,
the testing of GGMs in the traveled section both in absolute/direct
and the orthometric height is measured from spirit leveling.
(i.e., against systematic offsets or biases) and in relative (i.e., geoid
Therefore, estimates of N h2H obtained for a number of individual
slopes) terms.
points spanning a region can be used to evaluate the quality of
N ggm derived from a gravitational model. Obviously, the dis-
crepancies in absolute (direct) geoid height differences DN, which
Theoretical Background are computed by

Global Geopotential Models DN ¼ N h2H 2 N ggm ¼ ðh 2 HÞ 2 N ggm ð4Þ

A GGM is used to determine the long-wavelength component of the represent errors in N ggm , as well as errors in h and H (Kotsakis and
Earth’s gravity field in the form of a set of normalized spherical Sideris 1999; Fotopoulos et al. 2003; Huang and Véronneau 2004).
coefficients Clm and Slm for degree l and order m. The GGM-derived However, despite the various error sources that contaminate GPS
geoid height to a maximum degree lmax is approximated by and spirit-leveling operations, GPS/leveling data (if properly ac-
(Heiskanen and Moritz 1967; Mainville et al. 1992) quired and processed) is generally more accurate than N ggm , and
lP
max P
l   therefore, it can be used as a quality check for global models (Pavlis
N ggm  R Plm ðsin wÞ Clm cos ml þ Slm sin ml ð1Þ et al. 2008). The standard validation procedure generally adopted
l¼2 m¼0 in the published literature relies on the computation and quality
analysis of DN-values obtained for a number of points of known
where R 5 mean Earth radius; (w, l) 5 location of the computation (measured) h and H spanning an area of interest. In our approach,
point; and Plm 5 fully normalized Legendre functions for degree l h-values were obtained from the geodetic surveying of the railway
and order m, computed by tracks using a custom-built multisensor mobile mapping system,
 0:5 whereas their corresponding H-values came from the railway
ðn 2 mÞ! construction plans. Based on h- and H-values, we estimate N h2H
Plm ðsin wÞ ¼ kð2n þ 1Þ Plm ðsin wÞ ð2Þ (hereafter denoted by N nav ) along the railway traverse, and
ðn þ mÞ!

with k 5 1 for M 5 0; and k 5 2 for m  0. The associated Legendre


functions Plm can be computed using recursive formulas such as
those found in Holmes and Featherstone (2002). Generally, the
higher the number of coefficients adopted in a model, the more
detailed the model is, because it contains shorter-wavelength in-
formation about Earth’s gravity field.
Since the release of EGM96, more than 25 new GGMs have
become available to the scientific community through the Inter-
national Centre for Global Earth Models (ICGEM). These models
have led to significant improvement of our knowledge of the Earth’s
gravitational field and the geoid. However, the release of EGM08 by
the U.S. National Geospatial Intelligence Agency (NGA) was a
breakthrough in Earth’s gravity field mapping. As opposed to pre- Fig. 1. Relationships among ellipsoid (h), orthometric (H), and geoid
vious models, EGM08 is complete to spherical harmonic degree (N) heights and relative geoid height differences (dN)
and order 2,159. It also contains additional coefficients extending to

96 / JOURNAL OF SURVEYING ENGINEERING © ASCE / MAY 2013

J. Surv. Eng., 2013, 139(2): 95-104


subsequently, we compute the DN-values for both EGM96 and results obtained are discussed and evaluated along a test line in
EGM08. The analysis of DN-values is important, as it can reveal Greece.
various error sources (Erol and Erol 2012) associated with the
vertical datum of an area (e.g., inconsistencies between the local Spirit Leveling and Rail-Track Survey Data
vertical datum and the geoid) and the construction of GGMs (e.g.,
zero- and first-degree bias terms and long-wavelength systematic
errors). Target Area
As opposed to the absolute validation of a GGM, relative (or
differential) testing is generally free of long-wavelength systematic The navigated traverse adopted in this study coincides with the
errors as well as zero- and first-order terms, and thus, it can be used to suburban main line track connecting the metropolis of Athens with
assess the accuracy of geoid gradients and to identify weaknesses in the north of Peloponnesus, Greece. The surveyed section connects
the model data. In this study, relative testing along the surveyed the Ska rail traffic center in Athens to the terminal station at Kiato via
Downloaded from ascelibrary.org by Universidad Nacional De Ingeniería on 12/28/20. Copyright ASCE. For personal use only; all rights reserved.

section was accomplished through the comparison of GGM and Corinth (Fig. 2). It includes 101 km of high-speed, electronically
the GPS/leveling geoid slopes. As shown in Fig. 1, for every sub- operated double tracks that serve nine stops. The route is equipped
sequent pair of points i and j lying on the surveyed traverse, the rel- for bidirectional running with track centers located 4.20 m apart at
ative geoid height differences dN are computed for N nav and N ggm , standard (1,435 mm) rail gauge.
respectively, by The Ska-Kiato railway was selected on the basis of the following
criteria. Firstly, accurate orthometric heights were made available
  from the construction (as-built) plans at regular intervals along the
dNijnav ¼ Njnav 2 Ninav ¼ hj 2 Hj 2 ðhi 2 Hi Þ ¼ dhij 2 dHij railway line. Secondly, geoid undulations in the test area exhibit an
ð5Þ unusually wide variation. As shown in Fig. 3, the railway track cuts
through the contour lines (isogeoid undulation lines) of the EGM08
map, resulting in a total geoid undulation difference on the order of
dNijggm ¼ Njggm 2 Niggm ð6Þ several meters over a distance of 101 km. As a result, in addition to
geoid testing in absolute terms, this traverse also offers an interesting
and finally, the discrepancies in relative geoid height differences way to study the relative geoid differences (geoid slope). Finally, it is
are computed by emphasized that the selected railway line lies in the vicinity of the
Gulf of Corinth, which is considered to be one of the most prominent
ddNij ¼ dNijnav 2 dNij
ggm
ð7Þ active rifts in southern Europe (Bell et al. 2011; Clarke et al. 1997;
Elias et al. 2009). For these reasons, this study is expected to
contribute to our knowledge concerning the geoid topography of
From Eqs. (5) and (6), it is evident that any errors that are common this region.
to both points i and j cancel on differencing. Therefore, in the rel-
ative testing of the geoid, ellipsoid height differences and spirit-leveled
orthometric height differences are used to evaluate the accuracy of Orthometric Heights
GGM-derived geoid gradients (Featherstone 2001). This quality check
is important for GPS users, as they aim to determine orthometric height Helmert-type orthometric heights were made available at 20-m
differences dHij derived by Eq. (5), the accuracy of which is dependent intervals along the rail tracks from the railway line maintenance
on the accuracy of dhij and, especially, dHij . Moreover, the dis- plans. In fact, the rail tracks were surveyed during and after con-
crepancies in relative geoid differences ddNhij , given by Eq. (7), struction by spirit leveling in closed-loop traverses. A threshold
provide a useful measure of the consistency between measured value to reject spirit-leveling measurements between consecutive
and GGM-derived geoid gradients. nodes was adopted
In the following sections, the field procedures, the data- pffiffiffi
processing steps adopted, and the comparative analyses of the sH ¼ 6 0:7 S mm ð8Þ

Fig. 2. Rail-track section of the suburban railway of Athens from Ska rail traffic center to the terminal station of Kiato

JOURNAL OF SURVEYING ENGINEERING © ASCE / MAY 2013 / 97

J. Surv. Eng., 2013, 139(2): 95-104


where S 5 length of the leveling line in kilometers. Final orthometric for this study was undertaken. Comparisons between the two surveys
heights were obtained using the observation-equations least-squares revealed small height differences (less than 3 mm; D. Bezas, personal
adjustment method (Cross 1983). The reference surface of these communication, January 12, 2010). Also, independent leveling ties
heights is defined by the mean sea level (period 1933–1978) at the undertaken by the authors at selected checkpoints along the surveyed
Hellenic vertical reference frame (HVRF) fundamental tide-gauge path verified these findings. In absolute terms (i.e., in relation to the
station located at the Piraeus port (Takos 1989). Notably, the equipotential surface of Earth’s gravity field), the accuracy of observed
orthometric heights provided refer to the top of the lower rail track. orthometric heights reflects the information used to define the
Therefore, to obtain orthometric heights at the single-track cen- reference surface of the HVRF (unknown gravity potential W0 ;
terline, original heights were corrected for superelevation (also Takos 1989) realized through the mean sea level at the Piraeus port.
referred to as rail cant).
The relative accuracy of the available orthometric heights is mainly
Ellipsoid Heights
driven by the accuracy of leveling measurements and their adjustment
Downloaded from ascelibrary.org by Universidad Nacional De Ingeniería on 12/28/20. Copyright ASCE. For personal use only; all rights reserved.

and is found to be on the order of 10 mm for the entire survey. Luckily, Multisensor Kinematic-Surveying System
because of contractual obligations, two independent leveling surveys Ellipsoid heights of the railway tracks were obtained using a suitably
were conducted within a 9-month period before the kinematic survey equipped rail-track survey vehicle (Fig. 4). The recording vehicle

Fig. 3. Geoid undulations for EGM08 derived with respect to WGS84 over Greece

Fig. 4. Mobile mapping system used for navigation data acquisition: (a) rail-track recording vehicle; (b) and (c) reduction of sensor measurements to
the top of rail; and (d)–(f) GNSS antennas, INS, and DMI sensors fixed on the roof, the running gear, and the rear wheel of the vehicle, respectively

98 / JOURNAL OF SURVEYING ENGINEERING © ASCE / MAY 2013

J. Surv. Eng., 2013, 139(2): 95-104


is a railway diesel coach that was made available by the Hellenic Data Acquisition and Processing
Railways Organization. It has a capacity of 20 passengers and
Vehicle-navigation data acquisition involved a two-way survey of
maximum operating speed of 100 km/h. The vehicle navigation
the railway line. To facilitate a direct comparison of the navigation
system employed three dual-frequency GPS receivers and a high-
solutions obtained in opposite directions, the railway coach ran on
accuracy inertial unit [inertial navigation system (INS)] aided by
the same railway track forward and backward, so that the same
a digital odometer [distance-measuring instrument (DMI)] for linear/
longitudinal traverse was surveyed twice. Also, to achieve optimal
chainage referencing (Table 1). To minimize the effects of bogie-
performance of the INS, no layovers were allowed, and the vehicle
induced vibrations, the INS was mounted directly on the running
traveled at an average running speed .90 km/h. This is particularly
gear (chassis) of the vehicle. Prior to data collection, a dimensional
important for sections of bad global navigation satellite system
survey was undertaken to compute the relative locations (lever arms)
(GNSS) signal reception such as in tunnels and steep cliffs along the
of each sensor, so that all measurements were reduced to a predefined
way. For this reason and to avoid high traffic volumes, the survey was
point on the recording vehicle. For this purpose, a local geodetic
Downloaded from ascelibrary.org by Universidad Nacional De Ingeniería on 12/28/20. Copyright ASCE. For personal use only; all rights reserved.

undertaken during the night. To ensure high-quality navigation data,


control network was established surrounding the railway vehicle as
three GNSS base stations were established along the railway corridor
shown in Fig. 5 and measured using a high-accuracy total station
at distances not exceeding 30 km apart, so that the maximum baseline
and a spirit level. Finally, to associate ellipsoid heights with the
length from the recording vehicle to the nearest GNSS base station
available orthometric heights, the rail-track vehicle locations were
was less than 15 km. The WGS84 coordinates of the three base
reduced to the top of the rail followed by the rail-track centerline
stations were computed from the Dionysos Satellite Observatory of
(Gikas and Daskalakis 2008; Gikas and Stratakos 2012).
the National Technical University of Athens (DION) permanently
operating GNSS station using the observation scheme shown in
Table 1. Railway Vehicle Mobile Mapping System Geodetic Sensors Fig. 6. The three GNSS base stations and the rover receiver were set
Sensor type Specifications
to record data at 5 Hz with INS corrections sampled at 200 Hz.
Data processing involved filtering the navigation data [GNSS/
GNSS base station 13 DL-V3 @ 5Hz inertial measurement unit (IMU)/DMI] using an advanced Kalman
23 Trimble 5800 @ 5Hz filtering algorithm (NovAtel 2010; Gikas et al. 2008) to produce the
GNSS rover 13 PROPAK-V3 @ 5Hz optimum blended navigation solution that minimizes the effects of
INS 13 iMAR-FSAS @ 200Hz GNSS signal multipath and shadows. Output data consisted of
Digital odometer 13 Corsys Darton WPT-M12 vehicle coordinates obtained along the surveyed section in WGS84
and their projected equivalents in the Hellenic Geodetic Reference
System 1987 (HGRS87). Also, vehicle kinematics (velocity, ac-
celeration) as well as additional quality measures were obtained.
Ellipsoid heights along the surveyed track were consistently com-
puted with respect to the WGS84 reference ellipsoid, starting from
their Cartesian Earth Centered Earth Fixed (ECEF) WGS84 coor-
dinates. The accuracy of the ellipsoid heights is estimated to vary
from 4 to 7 cm, whereas in tunnel and steep-slope sections, these
values increase locally by up to 50%. The analysis also confirmed
relatively small (∼2 cm) differences in ellipsoid heights obtained
between the two surveys performed in opposite directions.

Geoid Undulations

Kinematic-Survey-Derived Geoid Undulations

Fig. 5. Geodetic control network used to measure the lever arms among Navigation-based geoid undulations were computed along the sur-
sensor locations veyed railway section from Eq. (3). As shown in Fig. 7, the location of
H-values refers horizontally to the double-track centerline (Point 1)

Fig. 6. Baseline length and observation periods used to compute the location and ellipsoid heights of base stations A, B, and C; observations made from
DION permanently operating GNSS station; for validation purposes, the same baselines were reobserved for relatively longer observation periods and
revealed differences in base-station heights on the order of 5 mm; data processing was performed using Bernese GNSS Software

JOURNAL OF SURVEYING ENGINEERING © ASCE / MAY 2013 / 99

J. Surv. Eng., 2013, 139(2): 95-104


and vertically to the top of the lower rail (Point 2). In contrast, rail- Fig. 8 shows the longitudinal profile obtained for both the
reduced, kinematic-surveying h-values refer to the top of the left orthometric and ellipsoid heights. From this plot, it is evident that
(running from Kiato to Ska station) rail on the same single track (Point orthometric heights vary between 160 (Kiato station) and 10 m
3). Therefore, to reduce the orthometric and ellipsoid heights to (Ska station). As expected, ellipsoid heights follow the same trend;
a common datum (Point 4), rail-track alignment information (i.e., however, a mean height difference of nearly 35 m is observed, which
superelevation, direction of turn, and cross-section geometry) was represents the average geoid height value of the surveyed section.
accordingly applied to the H- and h-values. Also, given the high Fig. 9 shows the geoid undulations derived from the navigation/
sampling rate of the navigation system, h-values were desampled, so leveling measurements. In this plot, the gray solid line represents the
that every point of known H was paired with the h-value of its nearest N nav -values obtained directly by subtracting H from h. The observed
fixed point obtained from the navigation solution. fluctuations primarily indicate the weaknesses in the navigation
solution. Notably, maximum variations occur at obscured (tunnel or
steep-gradient) sections. To mitigate these effects, a moving-average
Downloaded from ascelibrary.org by Universidad Nacional De Ingeniería on 12/28/20. Copyright ASCE. For personal use only; all rights reserved.

filter was consistently applied to the raw N nav -estimates. The filter
parameters were selected in accordance with the accuracy of the
navigation solution. The resulting N nav -estimates are denoted by the
black dashed line in Fig. 9. Finally, to ensure that the filtered geoid
heights are free of errors, a limited number of static GPS/leveling
checks were undertaken at selected points (see Fig. 9) along the
railway corridor. A comparative analysis of the results obtained from
these tests with those derived from the kinematic survey indicated
Fig. 7. Reference points for measured orthometric and ellipsoid heights small (1–2 cm) differences.

Global Geopotential Model–Derived Geoid Undulations


Geoid undulations N ggm were computed at points lying every 20 m
along the surveyed railway track using EGM08 (degree 2,190 and
order 2,159). Also, for comparison purposes, geoid heights have been
computed at the same points for the older EGM96 model to degree and
order 360. The GGM-based geoid undulations have been computed
with respect to WGS84 through the h_synth_WGS84.exe and F477.
exe software programs that are freely provided by the NGA website
(Pavlis et al. 2008). Note that the computations for the spherical
harmonic synthesis of N ggm have been performed in the tide-free
system. Therefore, this software applies a constant, zero-degree term
(N0 5 241:0 cm) to geoid heights computed using EGM08. This
term denotes the contribution of the zero-degree harmonic N ggm with
Fig. 8. Orthometric (solid line) and ellipsoid (dashed line) height respect to WGS84 derived from a mean Earth ellipsoid for which the
profiles obtained along the surveyed section estimated parameters in the tide-free system are a 5 6;378;136:58 m
and 1=f 5 298:257686 (Pavlis et al. 2008).

Fig. 9. Raw (gray solid line) and filtered (black dashed line) geoid heights obtained along the surveyed section; vertical bars indicate tunnel and steep
slope sections; P1 through P4 denote the location of static GPS/leveling checkpoints

100 / JOURNAL OF SURVEYING ENGINEERING © ASCE / MAY 2013

J. Surv. Eng., 2013, 139(2): 95-104


Geoid Height Evaluation and Discussion Fig. 10 in more detail shows that the EGM08-derived geoid heights
exhibit a significant increase compared with their N nav equivalents.
This represents a discrepancy between the Hellenic vertical datum and
Evaluation of Absolute Geoid Heights the equipotential surface realized by the EGM08 model along the
Fig. 10 depicts the geoid undulations obtained along the traveled surveyed traverse. This difference is not visible for EGM96 values
section for the navigation/leveling-based and GGM-based solutions. because of its significantly lower-degree harmonics (nmax 5 360).
The three curves correspond to the filtered N nav and N ggm for EGM08 This demonstrates a significant improvement that EGM08 offers over
and EGM96. From this figure, a number of points are immediately the test area in comparison with that of the older EGM96. In fact, the
evident. The first thing to note is a gradual increase in geoid heights EGM08-based N-values run in parallel and show an almost identical
from west (Peloponnesus, Kiato station) to east (Athens, Ska station). trend with the N nav estimates. In contrast, the EGM96-based N-values
This is irrespective of which N-estimate is considered. In fact, a total exhibit a gradually decreasing separation compared with N nav , ranging
difference on the order of 4.6 m is observed over the 101-km distance from about 80 cm at Kiato to about 5 cm at Ska. These effects are
Downloaded from ascelibrary.org by Universidad Nacional De Ingeniería on 12/28/20. Copyright ASCE. For personal use only; all rights reserved.

traveled—representing a large local geoid gradient. Examination of also evident in Fig. 11 and Table 2.

Fig. 10. Geoid heights obtained along the surveyed section: N nav (solid line), N EGM08 (dashed line), and N EGM08 (dotted line)

Fig. 11. Absolute geoid height differences obtained along the surveyed section: N nav 2 N EGM08 (dashed line) and N nav 2 N EGM96 (dotted line)

JOURNAL OF SURVEYING ENGINEERING © ASCE / MAY 2013 / 101

J. Surv. Eng., 2013, 139(2): 95-104


Fig. 11 shows the absolute geoid height differences (DN) areal extent to a countrywide estimate based on a larger (1,542) and
computed for the EGM08- and EGM96-based geoid undulations uniformly distributed set of points.
(N ggm ) from their N nav equivalents, whereas Table 2 summarizes the Except for the systematic offset observed between N nav and N ggm ,
statistics of the linear trend analysis for the geoid height differences. an important feature to note from Fig. 11 and Table 2 is the SD and
From the results in Fig. 11 and Table 2, it is evident that an estimated tilt in DN-values. Compared with EGM96, the SD (1-s level) of the
bias of –31.4 cm contaminates the EGM08 DN-values. This dif- EGM08 geoid height differences decreases by a factor of nearly four
ference suggests that, along the tested traverse, the Hellenic vertical (i.e., from 622.6 to 65.9 cm), which is a significant improvement in
datum that is realized through the geoid heights N nav is located local geoid determination. Obviously, the high SD (622.6 cm) and
approximately 31 cm below the EGM08 WGS84 tide-free system. tilt (–7.3 ppm) obtained from the trend analysis for EGM96 reflect
The existence of such a systematic offset is likely associated its limited spatial resolution.
with the internal accuracy and consistency of the HVRF in the area,
as well as with long-/medium-wavelength errors in the original Evaluation of Relative Geoid Heights
Downloaded from ascelibrary.org by Universidad Nacional De Ingeniería on 12/28/20. Copyright ASCE. For personal use only; all rights reserved.

spherical harmonic coefficients of EGM08.


In a previous study of GGMs in Greece based on GPS/leveling Another method to ascertain the quality of a GGM model is in
data, Kotsakis and Katsambalos (2010) computed a bias of –37.7 a relative manner, that is, to compute and compare the GGM-derived
cm over the entire mainland of Greece. This value refers to the relative geoid height differences (dN) with those obtained from the
EGM08 GRS80 geoid using a zero-degree term value N0 5 measured (GPS/leveling) geoid undulations. Fig. 12 and Table 3
2 44:2 cm. This bias estimate, if transformed to a geoid undulation contain the relative geoid height differences and their associated
without any zero-degree term (i.e., with respect to an ideal ellipsoid statistics obtained from the measured and the modeled geoid heights
whose semimajor axis remains numerically unspecified), reads along the surveyed section. Note that this figure conveys useful
information, but its value is largely driven by the resolution (half-
 
DN ¼ N h2H 2 N ggm ¼ N h2H 2 N ideal þ N0 0N h2H 2 N ideal wavelength) of the global models and the accuracy of the navigation
solution. For this reason, relative geoid height differences were
¼ DN þ N0 ¼ 2 37:7 2 44:2 ¼ 2 81:9 cm computed using Eq. (6) consistently applied at neighboring points
ð9Þ spaced 2.5 km apart. Also, to make a comparison between the
relative geoid height differences, a five-point moving-average filter
Similarly, our estimated bias value if transformed in the same ideal was applied to the dN-values. The first thing to note from Fig. 12
ellipsoid reads and Table 3 is that all three dN-estimates exhibit similar variation
  patterns and statistical indicators. However, the smaller SD and the
DN ¼ N nav 2 N ggm ¼ N nav 2 N ideal þ N0 0N nav 2 N ideal increased discrepancies observed between the EGM96-based
dN-values and those obtained for the EGM08 and GPS/leveling
¼ DN þ N0 ¼ 2 31:4 2 41:0 ¼ 2 72:4 cm geoid heights relate to the lower spatial resolution of EGM96.
ð10Þ The statistics of the discrepancies of the relative geoid height
differences (ddN) are given in Table 4. As a comparison, Table 4
which implies that our bias estimate differs by 9.5 cm compared with contains ddN-values computed at two spatial scales, i.e., for
the estimate computed by Kotsakis and Katsambalos (2010). This neighboring points spaced 2.5 and 5 km apart. As expected, because
discrepancy, though it is seemingly large, can still be justified, as it of differencing in dN-values, computations result in relatively
results from the comparison of a single traverse with rather limited small means. Also, the increase in the SD for the EGM96-based
ddN-values between the denser (2.5-km) and the coarser (5-km) data
sample may partially reveal the lower spatial resolution of this
model; however, the major contribution to this effect is clearly the
Table 2. Linear Trend Analysis of the Absolute Geoid Height Differences
DN 5 N nav 2 N ggm Obtained for the EGM08-Based and the EGM96-Based
smaller amount of sample data associated with the longer (5-km)
Geoid Heights separation between subsequent points.
Statistic DN EGM08 DN EGM96
Conclusions
Mean (cm) 2310.4 2290.6
Tilt (ppm) 210.1 270.3
This paper described a multisensor surveying-engineering tech-
s (cm) 650.9 6220.6
nique for estimating the geoid heights along longitudinal railway
Minimum (cm) 2490.1 2810.6
traverse lines. The proposed approach relies on the kinematic sur-
Maximum (cm) 2210.4 20.0
veying of railway track lines of known orthometric heights. The

Fig. 12. Relative geoid height differences obtained along the surveyed section: dN nav (solid line), dN EGM08 (dashed line), and dN EGM08 (dotted line)

102 / JOURNAL OF SURVEYING ENGINEERING © ASCE / MAY 2013

J. Surv. Eng., 2013, 139(2): 95-104


Table 3. Statistics of the Relative Geoid Height Differences dNij 5 d 5 rail-track center clearance;
Nj 2 Ni Obtained for the Navigation/Leveling, EGM08-Based and EGM96- e 5 superelevation (rail cant);
Based Geoid Heights f 5 ellipsoid flattening;
Statistic (cm) dN nav dN EGM08 dN EGM96 g 5 rail gauge;
H 5 Helmert orthometric height;
Mean 2120.6 2120.1 2100.9
s 620.2 620.3 610.6
m, n 5 degree and order of normalized spherical
Minimum 2150.7 2160.7 2120.6 coefficients;
Maximum 280.6 280.4 260.8 N0 5 geoid height attributed to zero-degree
harmonic term;
Table 4. Statistics of the Discrepancies in Relative Geoid Height Plm 5 normalized Legendre functions;
Differences ddNij 5 dNijnav 2 dNijggm Obtained for the EGM08-Based and R 5 Earth radius;
5
Downloaded from ascelibrary.org by Universidad Nacional De Ingeniería on 12/28/20. Copyright ASCE. For personal use only; all rights reserved.

EGM96-Based Geoid Heights S length of leveling line;


EGM08 EGM96 W0 5 gravity potential;
DN 5 absolute geoid height difference;
ddN ddN ddN ddN dH 5 difference in orthometric height;
Statistic (cm) @ 2.5 km @ 5.0 km @ 2.5 km @ 5.0 km dh 5 difference in ellipsoid height;
Mean 20.2 20.2 21.6 22.7 dN 5 relative geoid height difference;
s 61.4 61.9 61.7 63.5 ddN 5 discrepancy in relative geoid height
Minimum 23.2 23.8 24.5 26.4 difference;
Maximum 2.5 3.2 2.3 4.3 l, w 5 geodetic latitude and longitude; and
s 5 SD.
main contribution of the method is its ability to measure ellipsoid
heights with relatively high accuracy in a short acquisition time, with
dense point spacing, and in an inexpensive manner. Therefore, it can
Subscripts
potentially be used to develop a set of geoid calibration lines and to
serve as a check against geopotential gravity models. Moreover, i, j 5 positive integer indices.
resurveying of the same traverses over time would yield a direct and
independent check of actual geoid change in the study area. These
quality checks, under certain circumstances, could stand as in situ
validations of models of geoid change coming from various sources References
such as mantle dynamics models. As a minimum prerequisite, the
method requires known orthometric heights along the surveyed Amos, M. J., and Featherstone, W. E. (2003). “Comparisons of recent global
lines, projection of sensor measurement on the rail tracks needs to geopotential models with terrestrial gravity field data over New Zealand
be carefully computed, and GNSS/INS data need to be thoroughly and Australia.” Geomat. Res. Australas., 79, 1–20.
Andritsanos, V. D., Fotiou, A., Paschalaki, E., Pikridas, C., Rossikopoulos,
acquired and processed. The geographic distribution of potential test
D., and Tziavos, I. N. (2000). “Local geoid computation and evaluation.”
lines in an area is totally constrained by the structure and quality of Phys. Chem. Earth, Part A Solid Earth Geod., 25(1), 63–69.
the existing railway network. Bell, R. E., McNeill, L. C., Henstock, T. J., and Bull, J. M. (2011).
Implementation of the method at a railway line in central Greece “Comparing extension on multiple time and depth scales in the Corinth
has demonstrated its potential both in terms of efficiency and in Rift, Central Greece.” Geophys. J. Int., 186(2), 463–470.
terms of the quality of the results obtained. Navigation data pro- Bernese GNSS Software 5.0 [Computer software]. Bern, Switzerland, As-
cessing resulted in ellipsoid heights with an accuracy of 4 to 7 cm for tronomical Institute, Univ. of Bern.
the majority of the test line. Analysis of the GGM verification results Clarke, P. J., et al. (1997). “Geodetic estimate of seismic hazard in the gulf of
testified to the superiority of EGM08 over the older EGM96 grav- Korinthos.” Geophys. Res. Lett., 24(11), 1303–1306.
itational model. Specifically, an improvement by a factor of four Cross, P. A. (1983). Advanced least squares applied to position fixing, North
East London Polytech, London.
(i.e., from 622.6 to 65.9 cm; 1-s level) was observed in the SD
Dawod, G. M., Mohamed, H., and Ismail, S. (2010). “Evaluation and ad-
between the EGM96-based and EGM08-based DN-residuals. Also, aptation of the EGM2008 geopotential model along the northern Nile
despite the moderate length of the tested traverse, relative geoid valley, Egypt: Case study.” J. Surv. Eng., 136(1), 36–37.
height analyses verified the higher spatial resolution of EGM08 Delikaraoglou, D., and Mintourakis, I. (2010). “On the merging of hetero-
against that of EGM96. geneous height data from SRTM, ICESat and survey control monuments
for establishing vertical control in Greece: An initial assessment and
Acknowledgments validation.” Gravity, Geoid and Earth Observation Int. Association of
Geodesy Symp., S. Mertikas, ed., Springer, Berlin, 289–294.
Elias, P., et al. (2009). “Permanent scatterer InSAR analysis and validation in
The authors acknowledge Dr. N. Pavlis, NGA Office of GEOINT the gulf of Corinth.” Sensors (Basel Switzerland), 9(1), 46–55.
Sciences, for instructive and helpful remarks and the Hellenic Rail- Ellmann, A. (2010). “Validation of the new earth gravitational model
ways Organization for providing the railway coach and suitable per- EGM08 over the Baltic countries.” Gravity, Geoid and Earth Obser-
sonnel to collect the field data. vation Int. Association of Geodesy Symp., S. Mertikas, ed., Springer,
Berlin, 489–496.
Erol, B. (2012). “Spectral evaluation of Earth geopotential models and an
Notation
experiment on its regional improvement for geoid modelling.” J. Earth
Syst. Sci., 121(3), 823–835.
The following symbols are used in this paper: Erol, B., and Erol, S. (2012). “GNSS in practical determination of regional
a 5 semimajor axis of ellipsoid; heights.” Global navigation satellite systems: Signal, theory and
Clm , Slm 5 normalized spherical coefficients; applications, J. Shuanggen, ed., InTech, Croatia, 127–160.

JOURNAL OF SURVEYING ENGINEERING © ASCE / MAY 2013 / 103

J. Surv. Eng., 2013, 139(2): 95-104


Erol, B., Sideris, M. G., and Çelik, R. N. (2009). “Comparison of global Kotsakis, C., Katsambalos, K., and Ampatzidis, D. (2011). “Estimation of
geopotential models from the CHAMP and GRACE missions for the zero-height geopotential level WoLVD in a local vertical datum from
regional geoid modeling in Turkey.” Stud. Geophys. Geod., 53(4), inversion of co-located GPS, leveling and geoid heights: A case study in
419–441. the Hellenic islands.” J. Geod., 86(6), 423–439.
European Space Agency. (2009). “Gravity field and steady-state ocean Kotsakis, C., Katsambalos, K., Ampatzidis, D., and Gianniou, M. (2010).
circulation explorer.” Æhttp://earth.esa.int/GOCE/æ (Jul. 30, 2011). “Evaluation of EGM08 using GPS and leveling heights in Greece.”
Featherstone, W. E. (2001). “Absolute and relative testing of gravimetric Gravity, Geoid and Earth Observation Int. Association of Geodesy
geoid models using global positioning system and orthometric height Symp., S. Mertikas, ed., Springer, Berlin, 481–488.
data.” Comput. Geosci., 27(7), 807–814. Kotsakis, C., and Sideris, M. G. (1999). “On the adjustment of combined
Fotiou, A., Grigoriadis, V. N., Pikridas, C., Rossikopoulos, D., Tziavos, I. N., GPS/leveling/geoid networks.” J. Geod., 73(8), 412–421.
and Vergos, G. S. (2010). “Combination schemes for local orthometric Mainville, A., Forsberg, R., and Sideris, M. G. (1992). “Global positioning
height determination from GPS measurements and gravity data.” system testing of geoids computed from geopotential models and local
Gravity, Geoid and Earth Observation Int. Association of Geodesy gravity data: A case study.” J. Geophys. Res., 97(B7), 11137–11147.
Downloaded from ascelibrary.org by Universidad Nacional De Ingeniería on 12/28/20. Copyright ASCE. For personal use only; all rights reserved.

Symp., S. Mertikas, ed., Springer, Berlin, 309–314. Martín, A., Anquela, A. B., Padin, J., and Berné, J. L. (2010). “Ability of
Fotopoulos, G., Kotsakis, C., and Sideris, M. (2003). “How accurately can the EGM2008 high degree geopotential model to calculate local geoid
we determine orthometric height differences from GPS and geoid data?” model in Valencia, Eastern Spain.” Stud. Geophys. Geod., 54(3), 347–366.
J. Surv. Eng., 129(1), 1–10. National Aeronautics and Space Administration. (2004). “Gravity recovery
GFZ German Research Centre for Geosciences. (2006). “The CHAMP and climate experiment.” http://www.csr.utexas.edu/grace/gravity/ (Jul.
mission.” Æhttp://www.gfz-potsdam.de/pb1/op/champ/results/grav/010_ 30, 2011).
eigenchamp03s.htmlæ (Jul. 30, 2011). NovAtel. (2005SPAN technology system user manual rev 7, NovAtel,
Gikas, V., and Daskalakis, S. (2008). “Determining rail track axis geometry Calgary, AB, Canada.
using satellite and terrestrial geodetic data.” Surv. Rev., 40(310), Pavlis, N. K., Holmes, S. A., Kenyon, S. C., and Factor, J. K. (2008). “An
392–405. earth gravitational model to degree 2160: EGM2008.” Proc., 2008 EGU
Gikas, V., Laflamme, C., Larouche, C., Kasapi, E., Soilemezoglou, G., and General Assembly, 13–18.
Paradissis, D. (2008). “Development of advanced positioning and Pavlis, N. K., Holmes, S. A., Kenyon, S. C., and Factor, J. K. (2012). “The
videometry tools for mobile mapping: Implementation in 1000 km of development and evaluation of the earth gravitational model 2008
roads in Greece.” Proc., 10th Int. Conf. on Applications of Advanced (EGM2008).” J. Geophys. Res., 117(B4), 2156–2202.
Technologies in Transportation, National Technical University, Athens, Rabah, M., and Kaloop, M. (2011). “The use of minimum curvature surface
Greece, 2863–2873. technique in geoid computation processing of Egypt.” Arab J. Geosci.,
Gikas, V., and Stratakos, J. (2012). “A novel geodetic engineering method 6(4), 1–10.
for accurate and automated road/railway centerline geometry extraction Sadiq, M., and Ahmad, Z. (2009). “On the selection of optimal global
based on the bearing diagram and fractal behavior.” IEEE Trans. Intell. geopotential model for geoid modelling: A case study in Pakistan.” Adv.
Transp. Syst., 13(1), 115–126. Space Res., 44(5), 627–639.
Heiskanen, W. A., and Moritz, H. (1967). Physical geodesy, Freeman, San Sideris, M. G., Mainville, A., and Forsberg, R. (1992). “Geoid testing using
Francisco. GPS and leveling (or GPS testing using leveling and the geoid?).” Aust.
Holmes, S. A., and Featherstone, W. E. (2002). “A unified approach to the J. Geod. Photogramm. Surv., 57, 62–77.
Clenshaw summation and the recursive computation of very high degree Takos, I. (1989). “Adjustment of geodetic networks in Greece.” Bull.
and order normalised associated Legendre functions.” J. Geod., 76(5), Hellenic Mil. Geogr. Serv., 136, 19–93.
279–299. Tziavos, I., Vergos, G., and Grigoriadis, V. (2010). “Investigation of to-
Huang, J., and Véronneau, M. (2004). “GPS-leveling and CHAMP&- pographic reductions and aliasing effects on gravity and the geoid over
GRACE geoid models.” Proc., Joint CHAMP/GRACE Science Meeting, Greece based on various digital terrain models.” Surv. Geophys., 31(1),
GFZ German Research Centre for Geosciences, Potsdam, Germany. 23–67.
Kenyon, S., Factor, J., Pavlis, N., and Holmes, S. (2007). “Towards the next Tziavos, I. N., Vergos, G. S., Grigoriadis, V. N., and Andritsanos, V. D.
earth gravitational model.” Proc., 2007 SEG Annual Meeting, Society of (2012). “Adjustment of collocated GPS, geoid and orthometric height
Exploration Geophysicists, Tulsa, OK. observations in Greece. Geoid or orthometric height improvement?”
Kiamehr, R., and Sjöberg, L. E. (2005). “Comparison of the qualities of Geodesy for Planet Earth Int. Association of Geodesy Symp., Springer,
recent global and local gravimetric geoid models in Iran.” Stud. Geophys. Berlin, 481–488.
Geod., 49(3), 289–304. Ustun, A., and Demirel, H. (2006). “Long-range geoid testing by GPS-
Kotsakis, C., and Katsambalos, K. (2010). “Quality analysis of global leveling data in Turkey.” J. Surv. Eng., 132(1), 15–23.
geopotential models at 1542 GPS/levelling benchmarks over the Hel- Vanícek, P., and Martinec, Z. (1994). “The Stokes-Helmert scheme for the
lenic mainland.” Surv. Rev, 42(318), 327–344. evaluation of a precise geoid.” Manuscr. Geod., 19, 119–128.

104 / JOURNAL OF SURVEYING ENGINEERING © ASCE / MAY 2013

J. Surv. Eng., 2013, 139(2): 95-104

You might also like