You are on page 1of 17

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/278690726

Geodetic World Height System Unification

Chapter · January 2014


DOI: 10.1007/978-3-642-27793-1_83-1

CITATIONS READS

3 1,716

1 author:

Michael G. Sideris
The University of Calgary
270 PUBLICATIONS   3,223 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Selective de-correlation of GRACE harmonic coefficients View project

GIA with Non-linear or Composite Rheology View project

All content following this page was uploaded by Michael G. Sideris on 30 November 2016.

The user has requested enhancement of the downloaded file.


Handbook of Geomathematics
DOI 10.1007/978-3-642-27793-1_83-1
© Springer-Verlag Berlin Heidelberg 2014

Geodetic World Height System Unification


Michael Sideris
Department of Geomatics Engineering, University of Calgary, Calgary, AB, Canada

Abstract

Elevations are one of the positional attributes embedded in all geospatial data. They are essential
for a wide range of engineering and scientific activities. Some of the activities requiring precise
elevations are activities of high societal impacts, such as sea level rise, storm surges and coastal
inundation, floods and evacuation route planning, and crustal motion, subsidence, and other surface
deformations due to seismic, mining, or other events. In order to successfully monitor and manage
such events regionally (e.g., floods) or globally (e.g., sea level rise), the elevation information needs
to not only be accurate but to also refer to the same zero-height reference surface (vertical datum).
Accurate elevations can be obtained using spirit leveling, or by combining Global Navigation
Satellite Systems (GNSS) positioning methods with a model of the geoid (the equipotential surface
of the Earth’s gravity field approximating the idealized mean sea surface). So although the high
accuracy requirements can in general be met, the need for a common reference surface often
cannot. This is because there are currently over 100 different vertical datums around the world.
Their unification is therefore a scientific problem of high practical significance.
Height system unification is the process of determining the offsets between existing vertical
datums, each of which is defined with a fundamental zero-elevation point and the equipotential
(level) surface of the Earth’s gravity field that passes through that point. These offsets can be
estimated if the geodetic (from GNSS), physical (from spirit leveling), and geoid (from gravity
information) heights are known for the fundamental and/or other points in each datum. GNSS-
surveyed leveling benchmarks on land, GNSS-surveyed tide gauge stations at the coasts, and a
mean dynamic topography model at sea can be used as data, together with a geoid model. The need
to estimate geoid undulations implies the solution of a geodetic boundary value problem (GBVP);
see, e.g., Colombo (1980), Rummel and Teunissen (1988), Rapp and Balasubramania (1992),
Sansò and Venuti (2002), and Moritz (2013). Although normal heights and height anomalies
from the solution of Molodensky’s GBVP can and have been used for datum unification, here we
work with orthometric heights and geoid undulations from the solution of the classical (Stokes)
GBVP. We first review briefly the key concepts of leveling and vertical datum definition and
realization, and then discuss, in spherical approximation, the single (global) vertical datum GBVP
and the multiple vertical datum GBVP following primarily the developments of Rummel and
Teunissen (1988). We also introduce the functional and stochastic models used for the least-squares
estimation of datum offsets and their accuracies. Using an example from North America with
tide gauges, we address a few important practical aspects, such as the magnitude of the so-called
indirect bias term resulting from the use of biased local gravity data, and the commission and
omission geoid undulation errors. Other issues related to the need for common epoch, tide system,
and reference frame for all the data involved in the computations are also briefly addressed.


E-mail: sideris@ucalgary.ca

Page 1 of 16
Handbook of Geomathematics
DOI 10.1007/978-3-642-27793-1_83-1
© Springer-Verlag Berlin Heidelberg 2014

1 Introduction
Traditionally, elevations (actually, elevation differences) are obtained using spirit leveling mea-
surements between benchmarks anchored to the ground, supplemented by gravity observations.
This process is labor-intensive, time-consuming, terrain- and weather-dependent, and expensive.
To overcome these limitations, elevations are now also obtained by GNSS positioning methods
and the geoid; see Fig. 1. These two methods, however, do not produce the same type of heights.
The former provides elevations above the surface of a reference ellipsoid (ellipsoidal or geometric
heights h; length of line PQ in Fig. 2) while the latter provides heights above a realization of the
mean sea level (orthometric heights H ; length of curved line PPo in Fig. 2), which is represented by
the surface of the geoid. Such ellipsoidal heights have only a geometric meaning while orthometric
heights have a physical meaning as they indicate potential differences between stations; see
Sect. 2 below. They are therefore indispensable for engineering projects involving water storage
or transportation such as drainage, irrigation, dams, etc., and for environmental monitoring
applications such as sea level rise, coastal inundation, and floods. Thus, although GNSS elevations
can now be easily determined in places where traditional methods are impractical, these elevations
provide no information about potential differences and therefore are not useful for the types
of applications described above. Fortunately, it is possible to convert GNSS-derived ellipsoidal
heights h to orthometric heights H; through the use of geoid heights (or geoid undulations) N .
Because the deflection of the vertical angle between lines PQ and PPo in Fig. 2 is very small
(typically not exceeding one arc second), we can assume that points P , Q and Po are collinear and
therefore we can obtain orthometric heights as H D h  N .
The proliferation of GNSS positioning instrumentation and methods and the availability of
very precise models for the geoid enabled by dedicated gravity satellite missions such as the
Gravity Recovery and Climate Experiment (GRACE) and Gravity Field and Steady-State Ocean
Circulation Explorer (GOCE), as well as the very high cost of establishing and maintaining
traditional leveling benchmarks, have led many countries to consider the adoption of geoid-
based vertical datums and GNSS-based leveling techniques to replace their existing leveling-based
vertical datums. These national height modernization efforts make a national geoid model freely
available and hence allow users to “sample” the new datum at any point by simply using a GNSS
receiver. This has, on the one hand, led to unprecedented improvements in the efficiency, and thus

Fig. 1 Traditional and modern way of elevation determination (after Hofmann-Wellenhof and Moritz 2006)

Page 2 of 16
Handbook of Geomathematics
DOI 10.1007/978-3-642-27793-1_83-1
© Springer-Verlag Berlin Heidelberg 2014

topographic
H surface
h

Po geoid Ω
N
Q
ellipsoid E

Fig. 2 Geometry between ellipsoidal, orthometric, and geoid heights

the cost of leveling operations but, on the other hand, has made the need for unifying the various
existing height systems even more pressing.

2 Vertical Datum Definition and Realization


A vertical datum is the reference surface to which heights refer, i.e., it is the surface of zero
elevation in a particular height system. As mentioned in Sect. 1, for example, the heights that are
derived from GNSS observations are “measured” along the normal to the surface of a geocentric
ellipsoid. Thus, these ellipsoidal or geodetic heights h have as datum the ellipsoidal surface.
However, because ellipsoidal heights are geometric quantities with no physical meaning, other
height systems are in use, in which differences in elevations represent potential differences. The
defining equation for such a system is (Hofmann-Wellenhof and Moritz 2006)

Zh0
C Wo  WP 1
h0 D D ; gQ D 0 gdh0 (1)
gQ gQ h
0

where h0 denotes a generic height, the geopotential number C is the difference between the gravity
potential Wo of the datum and the gravity potential W of any point on the Earth’s surface (see
also Fig. 2), and gQ denotes the average gravity from the reference surface to the point on the
Earth’s surface along the line perpendicular to the gravity equipotential surfaces (plumb line). The
particular choice of gQ results in different height systems, namely, the orthometric (h0 D H when
gQ D g/,N normal (h0 D H  when gQ D N /, and dynamic (h0 D H d when gQ D D45ı / systems,
where  is the gravity of the reference ellipsoid,  is the geodetic latitude, and the overbar denotes
average gravity along the corresponding plumb line. A very good introduction to these height
systems and vertical datums, including tidal effects, is given in Jekeli (2000).
It is clear from Eq. (1) that although physical heights are not directly measurable (and neither is
W itself), they could be derived from potential differences. Rearranging Eq. (1)

Zh0
C D Wo  WP D gd h0 (2)
0

Page 3 of 16
Handbook of Geomathematics
DOI 10.1007/978-3-642-27793-1_83-1
© Springer-Verlag Berlin Heidelberg 2014

we see that in principle potential differences can be computed by measuring height differences
and gravity, which is feasible by use of leveling equipment with gravity meters. Leveled height
differences can be converted to dynamic, orthometric, or normal by applying small corrections
that involve gravity measurements; see Hofmann-Wellenhof and Moritz (2006), sections 4.2 to
4.4. We also see that to get the absolute potential of any point we must know the value Wo of the
datum (taken here as one of the level or equipotential surfaces of the Earth’s gravity potential), and
define a fundamental point on the datum surface where we set h0 D 0 in order to physically realize
the datum. For the sake of simplicity, from this point on we will use the orthometric height system.
National datums are typically realized by selecting as their fundamental point Po a coastal tide
gauge and setting HPo D 0, WPo D Wo , and then connecting the country’s leveling network to
this fundamental point. This traditional, though arbitrary, choice is due to the historical preference
of measuring heights above mean sea level (MSL) and the use of the geoid (that Wo equipotential
surface that best approximates the idealized MSL) as the vertical datum for orthometric heights
H . Obviously, if the fundamental points of two datums were on different equipotential surfaces
but with known potential difference, then we would be able to connect one datum to the other.
And if this difference were zero, then we would only have a single datum, which could be, for
example, a global geoid. If, however, they would each refer to the MSL as measured by a regional
or national tide gauge, then due to oceanographic and other effects they would not refer to the same
equipotential surface, and therefore it would be difficult to connect them unless we could model
the discrepancies between the local MSL and the geoid with sufficient accuracy.
Proper vertical datum unification by use of the MSL is in fact possible and relies on the
connection of the MSL at various tide gauges by a mean dynamic topography (MDT) model,
corresponding to H on land. MDT can be estimated either as the difference between the mean
sea surface heights derived from satellite altimetry, corresponding to h on land, and the geoid
undulation N , or from an ocean circulation model. This method, referred to as the ocean method
or the ocean leveling method (see, e.g., Cartwright and Crease 1963; Fischer 1977; Rummel and
Ilk 1995; Woodworth et al. 2012), will not be discussed here. Instead, we will concentrate on the
geodetic method that uses as observables orthometric heights from leveling, gravity measurements
and ellipsoidal heights from GNSS, and determines the geoid by solving the geodetic boundary
value problem (GBVP). A survey of strategies for solving the vertical datum problem using
terrestrial and satellite geodetic data can be found in Heck and Rummel (1990).

3 Single Vertical Datum GBVP


We assume that the geoid is the boundary surface  enclosing all masses (this can be achieved
by applying topographic and atmospheric corrections that account for the masses outside /.
Then, we linearize the gravity potential and the gravity value by approximating them by the
corresponding normal potential U and normal gravity  of a reference ellipsoid with the same
mass and rotation rate as the Earth; see also Section 4.1 and Table 1 in Moritz (2013). The surface
of the geoid  will also be approximated by the surface of the ellipsoid E, which are separated by
the geoid undulation N as in Fig. 3:

W DU CT (3)

Page 4 of 16
Handbook of Geomathematics
DOI 10.1007/978-3-642-27793-1_83-1
© Springer-Verlag Berlin Heidelberg 2014

n n′

P geoid
W = W0

Ω
N
gP

E Q
reference
ellipsoid
U = U ¢0
gQ

Fig. 3 Geometry of GBVP linearization

g D  C ıg (4)

H DhN (5)

Because U and  can be computed analytically, the problem of determining W is now reduced
to the problem of estimating the disturbing potential T , which, under the above assumptions, is a
harmonic function outside the masses and regular at infinity. ıg is called the gravity disturbance
and is the difference between g and  at the same point. Linearization of W and g using first-order
Taylor expansions gives the “corrections” to U and  as follows:
ˇ !
@U ˇˇ  
TP D WP  UP D WP  UQ C ˇ N D W P  UQ  N D Wo  .Uo  N /
@n0 Q (6)
D Wo C N
ˇ  ˇ   ˇ 
@W ˇˇ @U ˇˇ @W ˇ
ˇ @T @T
ıg D gP  P D  ˇ   0ˇ D P  ˇ D D P  (7)
@n P @n P @n P @n @h

Here, we have neglected the deflection of the vertical (the angle between the actual and normal
gravity vectors; see Fig. 3) and have approximated the vertical derivatives along the actual and
normal plumb lines by the derivative with respect to the normal to the ellipsoid.
Equation (6) provides a way of estimating the geoid undulation from T called Bruns’s formula

T Wo
N D  (8)
 

and Eq. (7) provides a way of relating the unknown T to the data ıg on the boundary surface.
Even in the current era of GNSS positioning, most countries do not store their gravity data as
gravity disturbances but as gravity anomalies g, which are the difference between gravity on the
geoid and normal gravity on the ellipsoid (at points P and Q, respectively, in Fig. 3). Therefore,
we will relate T to g and work with anomalies instead of disturbances. Starting from Eq. (7), we
can obtain the equation (boundary condition) that relates T to g on  as follows:

Page 5 of 16
Handbook of Geomathematics
DOI 10.1007/978-3-642-27793-1_83-1
© Springer-Verlag Berlin Heidelberg 2014

   
@T @ @ T Wo
 D ıg D gP  P D gP  Q C N D g  
@h @h @h   (9)
@T 1 @ 1 @
) C T D g C Wo
@h  @h  @h

We are now ready to define the GBVP by combining Eq. (9) with Laplace’s equation that expresses
that fact that T is a harmonic function outside . We will employ spherical approximation, by use
of sphere with radius r D R, where R is the mean radius of the Earth. In this case

1 @ 1 @ 2
D
P D (10)
 @h  @r R

and the single (global) vertical datum GBVP is then defined as follows:

r 2T D 0 outside the boundary


@T 2 2 (11)
  T C Wo D g on the boundary
@r R R
The datum problem can be solved for the potential upon introducing the regularity condition
T ! 0, when the geocentric distance r ! 1. For a computational point P , the solution for
T is given (spherical approximation, spherical boundary surface) by Stokes’s equation
ZZ
R
TP D S t . /g d D SP g (12)
4


X
1
2n C 1
St. / D Pn .cos / (13)
nD2
n1

where  is the unit sphere, is the spherical distance between data and computation point, Pn
is the Legendre polynomial of degree n, and St( ) is the Stokes function (see also Moritz 2013,
eq. 71). Then the geoid undulation can be obtained from Bruns’s equation (8):

Wo TP 1
NP D  C D No C SP g (14)
  

Although not necessary, the constant zero-degree geoid term

Wo
No D  (15)


can vanish by selecting Uo D Wo . Also note that the operator applied to T in Eq. (9) is invariant
with respect to first-degree harmonics but this singularity is avoided by locating the origin of the
coordinate system at the geocenter.

Page 6 of 16
Handbook of Geomathematics
DOI 10.1007/978-3-642-27793-1_83-1
© Springer-Verlag Berlin Heidelberg 2014

Fig. 4 Geometry of multi-datum GBVP (figure by L. Sanchez, modified by the author)

4 Multiple Vertical Datums GBVP


We turn now to the much more realistic problem of multiple, say J +1, datums. Following Rummel
and Teunissen (1988), we assume that the whole sphere  is covered by J +1 non-overlapping
vertical datum zones j , j D 0; 1; 2; : : :; J . Each of these vertical datums is defined by the level
surface with potential Woj passing through a fundamental point Poj ; see Fig. 4. The zone o defined
by Woo D Wo and Poo D Po is chosen arbitrarily as the reference datum with respect to which the
geopotential differences ıWoj

ıWoj D Wo  Woj (16)

and the associated vertical datum offsets will be computed. The single vertical datum problem is
thereby transformed to a multiple vertical datum problem with J unknown vertical datum offsets.
With respect to the corresponding values for the reference datum, in datum j not only the geoid
undulations N j and the orthometric heights H j will be biased but so will the gravity anomalies
g j , as well, as we have

Woj D Woj  Uo D Woj  Uo C Wo  Wo D Wo  ıWoj (17)

2
g j D g  ıW j (18)
R o
where, for simplicity, we assume the same reference ellipsoid for all datums. The more general
case of different datums having different reference ellipsoids and in ellipsoidal approximation has
been treated in Sansò and Usai (1995) and in Sansò and Venuti (2002). Equation (18) indicates
that the g j gravity anomalies are biased with respect to those on the reference datum by the
term 2ıWoj =R, which can be considered as the “free-air” reduction needed to reduce the gravity
anomaly g j from the zero-height level Woj to the reference Wo level.
We will now set up the GBVP for J + 1 datum zones. By use of the above equations, the GBVP
in Eq. (11) takes the form

r 2T D 0 outside the boundary


@T 2 2 2 j P
J
(19)
  T C Wo D g  cP ıWoj on the boundary
@r R R R j D1

Page 7 of 16
Handbook of Geomathematics
DOI 10.1007/978-3-642-27793-1_83-1
© Springer-Verlag Berlin Heidelberg 2014

j
where cP indicates whether the computation point belongs to zone j or not:

j 1; if P 2 j
cP D (20)
0; if P … j

The solution of the GBVP in Eq. (19) is


0 1
 j 2 X J
2 X
J
j @
TP D SP g D SP g 
j
c ıW jA
D SP g 
j j
ıW j I D TP  ıTP (21)
R j D1 P o R j D1 o P

ZZ
j R
IP D St . /d (22)
4
j

From Eqs. (21) and (22) we can obtain the expression for TP when computed from biased
anomalies as

 j 2 X
J
j j j
TP D TP C ıTP D SP g C ıWoj IP (23)
R j D1

Substituting the above equation into Bruns’s equation we obtain

Woj TP Wo ıWoj 1   2 P J


j
NP D  C D C C SP g j C ıWoj IP
     R j D1
j j (24)
T ıT
D No C ıN j C P C P
 

We see from Eq. (24) that the use of biased gravity anomalies in the solution of the GBVP
introduces two effects in the derived geoid: a direct bias term on N , i.e., the vertical datum offset
ıN j (the second term in the RHS of the equation)

ıWoj
ıN j D (25)


and an indirect bias term on T (the last term is the equation), which can be written explicitly as

2 X
j J
j;ind ıTP 2 2 j j
ıNP D D SP ıWo D ıN IP C
j
ıN i IPi (26)
 R R R
iD1;i¤j

Equation (26) shows that the geoid height of point P in the datum zone j is affected by the offsets
ıWoj of all datum zones j D 0; 1; : : :; J .

Page 8 of 16
Handbook of Geomathematics
DOI 10.1007/978-3-642-27793-1_83-1
© Springer-Verlag Berlin Heidelberg 2014

5 Computation of the Terms in the Multi-datum GBVP Solution


Before giving the general procedure for estimating the unknown datum offsets, it is worth
discussing the computation of the quantities that are derived from the biased gravity anomalies.
The term NP D  1 TP D  1 SP g j in Eq. (24) is not computed by global integration of gravity
j j

anomalies. Instead, the remove-restore technique is used in order to limit the integration within
a manageable spherical cap o around the computation point. First, the long wavelengths of the
gravity field are removed from the data by use of geopotential model (GM), then the residual
anomalies are the input to Stokes’s integral to produce residual undulations, and then the long
wavelength part of the geoid is computed from the same GM and restored. This procedure is
described by the following equations:

X
nmax X
n
 
gP;GM D  .n  1/ CN nm cos mP C SNnm sin mP PNnm .cos P / (27)
nD2 mD0

j
gres D g j  gGM (28)

j 1
NP;res D j
SP;res gres (29)


X
nmax X
n
 
NP;GM DR CN nm cos mP C SNnm sin mP PNnm .cos P / (30)
nD2 mD0

j 1 j
NP D SP g j D NP;GM C NP;res (31)


where CN nm , SNnm are the fully normalized spherical harmonic coefficients for T of degree n
and order m, PNnm are the fully normalized associated Legendre functions, and , are the geodetic
co-latitude and longitude of the computational point.
Typically, and especially in areas of rough topography, the remove-restore technique involves
also the removal of the topographic attraction from the gravity anomalies via a terrain reduction
such as, e.g., Helmert’s second condensation or the Residual Terrain Model (RTM) reduction,
and the restoration of the indirect effect of the particular topographic reduction used on the geoid
undulations. Because here we have assumed that the gravity anomalies have already been reduced
onto the boundary surface, we do not provide the equations for these reductions. The interested
reader should consult Sansò and Sideris (2013), Chapters 4 and 8, for detailed procedures and
formulas.
Note that in Eq. (29) the residual Stokes kernel

X
nmax
2n C 1
Stres . / D St. /  Pn .cos / (32)
nD2
n1

is used, where nmax is the maximum degree and order of the GM used in Eqs. (27) and (30). The
residual Stokes kernel should then also be used in Eq. (26) for the computation of the indirect

Page 9 of 16
Handbook of Geomathematics
DOI 10.1007/978-3-642-27793-1_83-1
© Springer-Verlag Berlin Heidelberg 2014

effect, which can be explicitly written as


j
j;ind ıTP 2 1
ıNP D D SP;res ıWoj D ıWoj s s Stres . /d j
 R 2 j
X 1
C ıWoi s s Stres . /d i (33)
2 i
iD1;i¤j

Equation (24) can now take the form


j j j;ind
NP D No C ıN j C NP;GM C NP;res C ıNP (34)

It is clear from Eq. (33) that the indirect bias term vanishes only if T is derived from unbiased
gravity anomalies. One way to do this is to compute g from a set of GM coefficients (see
Eq. 27), which have been derived purely from satellite methods, i.e., without any terrestrial
observations. Such GM models are the recent satellite-only GOCE-based GMs. Because of their
limited resolution, such models have to be supplemented above their nmax by either terrestrial data
or by coefficients from a high degree and order GM such as the EGM2008 model (Pavlis et al.
2012), because the residual (omission) part’s contribution to the value of N is typically a few
decimeters, and therefore cannot be neglected even though averaging can bring it to the sub-dm
j
level; see results for NP;res at tide gauges in North America in Table 1. However, because of the
use of the residual Stokes kernel in Eq. (33), the indirect term can safely be omitted. To illustrate
this we provide an example from Rangelova et al. (2014), where we used the four datum offsets
j;ind
that are given in Fig. 5 and computed ıNP for the case that the GO_CONS_GCF_2_TIM_R4 (or
TIM4 for brevity; Pail et al. 2011) GOCE-only model was truncated to various maximum degrees
and supplemented by local gravity anomalies.
The results in Fig. 6 show that although when the unmodified Stokes kernel is used the indirect
bias term exceeds 40 cm, this term drops to sub-cm level when the residual kernel corresponding
to the nmax used for the GM-contribution is used (nmax D 180 was used in this case as the GM error
increased when TIM4 was used with higher maximum degrees; see Amjadiparvar et al. 2013). In
fact for any nmax  70, the indirect effect was found to be below 1 cm. Similar results have also
been reported by Gerlach and Rummel (2013) and Gatti et al. (2013). It can thus be concluded
that when satellite-only GMs with nmax  70 are used to provide the long wavelength part of the
solution, the indirect term can safely be neglected for cm-level of accuracy.

j
Table 1 Statistics of the residual term NP;res computed from EGM2008 for 181  n  2190 at tide gauges
Region (# of TGs) Mean (cm)  (cm) Min (cm) Max (cm)
Canada Atlantic (7) 1 33 56 38
US Atlantic (28) 3 37 84 72
Canada Pacific (5) 6 40 36 64
US Pacific (17) 7 33 49 74
Gulf of Mexico (13) 3 26 62 28

Page 10 of 16
Handbook of Geomathematics
DOI 10.1007/978-3-642-27793-1_83-1
© Springer-Verlag Berlin Heidelberg 2014

Fig. 5 North American vertical datum offsets from the level surface with Wo = 62636856.0 m 2 s 2

Fig. 6 Indirect bias term with unmodified Stokes kernel and residual Stokes kernel for nmax = 180

6 Determination of the Datum Offsets and Their Accuracy


j
Using NP from Eq. (34) into Eq. (5) leads to the following observation equation for each point P
j
with given ellipsoidal height hP determined from GNSS positioning and orthometric height HP
from leveling in the datum j :
 
2 X i
J
j j j 2 j
lP D hP  HP  NP;GM  NP;res D No C 1 C IP ıN C
j
IP ıN i C vP (35)
R R
iD1;i¤j

where lP denotes the “observable” and vP denotes the stochastic residual. To solve for all J C 1
unknown offsets, namely, No and ıNj , j D 1; 2; : : :; J , at least one data point P per datum zone
should be given. Naturally, when more points are available, the solution is obtained by least-squares
adjustment. In the simplest case, this data point could be the fundamental tide gauge (TG) station of
the respective vertical datum (see Fig. 7), with known ellipsoidal height of the local MSL h D hMSL
j
at the GNSS-surveyed fundamental tide gauge station and height of the local MSL H D HMSL
above the national datum j . hMSL at each tide gauge is computed with respect to the reference
ellipsoid by reducing the ellipsoidal height of the tide gauge benchmark with the height difference
between the benchmark and the chart datum obtained by precise leveling and adding the measured
water level from the chart datum. The MSL values should be corrected for the inverse barometer
and nodal tide (Woodworth et al. 2012). Ideally, tide gauge stations with 19-year long records of
water levels should be used so that nodal tides, atmospheric pressure and storm events are averaged
out (Pugh 1987).
For a set of n  J C 1 observation equations, the matrix form of the functional model
of Eq. (35) is

Page 11 of 16
Handbook of Geomathematics
DOI 10.1007/978-3-642-27793-1_83-1
© Springer-Verlag Berlin Heidelberg 2014

Fig. 7 Geometry of determination of vertical datum offsets using tide gauges

AıN C v D l (36)

and the corresponding stochastic model is


˚ 
E fvg D 0; E vvT D Cl D Ch C CH C CNGM C CNres (37)

where E{} is the expectation operator, A is the design matrix, and C are variance-covariance
matrices (VCMs). The least-squares solution gives the vector of the datum offsets and their
variance-covariance matrix:
   1
O D AT C1 A 1 AT C1 l;
ıN Cı NO D AT C1
l l l A (38)

The above is the general least-squares solution that can be applied to the global unification of
height systems. It can also be used with, besides tide gauges, observations form GNSS/leveling
benchmarks and from ocean leveling, either separately for each data group or in a combined fashion
adjusting together all available data groups.
In practice, because of the negligible magnitude of the indirect effect, the RHS of Eq. (35) can
be simplified to No + ıNJ C vP . Another simplification often applied in practice is the use of
diagonal or block-diagonal VCMs in Eq. (37), either because of lack on information or to improve
the efficiency of the computations, provided the resulting accuracy loss is acceptable. For example,
Gerlach and Fecher (2012) have shown that the full VCM of NGM obtained from the propagation
of the GM coefficient errors can be approximated without significant loss of accuracy by using
only elements of the dominant m-block structure of the VCM. Following this approach, Rangelova
et al. (2014) found that the TIM4 commission error has a noteworthy north-south variation at tide
gauges in North America, varying from 1.9 cm at the northernmost tide gauges to 2.7 cm at the
southernmost tide gauges (see Fig. 8), while a more substantial range of 1.3 cm was seen when a
diagonal VCM was used. It must be mentioned here that the full geoid commission error must also
include the effects of the errors in gravity and topography data, which is not an easy task due to
the scarcity and questionable reliability of the available error information of such data. For a full
investigation on the quality of the least-squares solution, Xu and Rummel (1991) and Xu (1992)
should be consulted.
An often used simplification of Eq. (35) is to set No D 0. This is equivalent to assuming a
conventional value for Wo , either by selecting it to be equal to Uo or by computing it with data at
tide gauges or in the oceans. The geodetic/oceanographic approach to Wo , for example, relies on the
Gauss-Listing definition of the geoid, namely, that it is the level surface around which the average
permanent sea surface topography (SST) heights, Hs , are near zero when averaged globally over

Page 12 of 16
Handbook of Geomathematics
DOI 10.1007/978-3-642-27793-1_83-1
© Springer-Verlag Berlin Heidelberg 2014

Fig. 8 Commission geoid error from TIM4 at North American tide gauges

all oceanic regions Ss . This condition can be expressed mathematically as


 2
@ @ Wo  WP
s s HS2P d  ss d D 0 (39)
@Wo SS @Wo SS P

from which the value of Wo can be obtained as

WP 1
Wo D s s d= s s d (40)
SS P2 SS P
2

The SST is typically derived from altimetry over the oceans and Wo can be estimated from
gravity data. There are many problems to be resolved in this approach, including the spatial
coverage/integration area and choice of models (typically the polar regions are left out), the choice
of reference epoch and long-term variations of MSL (and, where significant, postglacial rebound,
PGR), referencing all data to the same epoch and tide system, and the choice (ideally, satellite-
only) and maximum degree of a GM. Although this approach has its proponents (see, e.g., Bursa
et al. 2008; Sanchez 2008), it has not been adopted here, on the one hand because of the inherent
problems just mentioned and, on the other hand, because selecting a particular Wo is not really
necessary for global height unification, unless governments decide to reference all their data to a
unified global vertical datum, which is very unlikely.
To conclude this section, we point out that besides knowing the error characteristics of all data
sets that enter into the least-squares adjustment, care must be taken so that all data are consistent
in terms of observation epoch (usually not the case), global terrestrial reference system (e.g.,
GNSS ellipsoidal heights should be given in a common International Terrestrial Reference Frame
and epoch), reference ellipsoid, height type, and permanent tide system (usually not the case as
coordinates are in the mean-tide system while gravimetric quantities are in the zero-tide system).
Some of these data consistency problems are discussed in Heck (2005). In areas with significant
crustal motion, heights of tide gauge benchmarks should be corrected using GNSS-derived vertical

Page 13 of 16
Handbook of Geomathematics
DOI 10.1007/978-3-642-27793-1_83-1
© Springer-Verlag Berlin Heidelberg 2014

crustal velocities or geophysical models of crustal motion due to PGR (Argus and Peltier 2010).
Such areas in North America are the Hudson Bay and the Atlantic coast of Canada, where some tide
gauge stations experience large subsidence due to the ongoing postglacial rebound in the region
and, as a result, the local mean sea level rises with respect to the crust (Rangelova et al. 2012).
In addition, water levels should be corrected for local long-term sea level changes determined at
the tide gauge stations. Such a correction is necessary in order to unify the epoch of the computed
MSL with the epoch of the ellipsoidal height of MSL.
Although no global vertical datum unification has been attempted yet, several investigations
containing actual results on the unification of regional and continental vertical datums have been
made. They can be found in articles by Featherstone and Filmer (2012) for Australia, Amos and
Feathersone (2009) and Tenzer et al. (2011) for New Zealand, Sanchez (2007) for South America,
and Sideris and Fotopoulos (2012) and Rummel et al. (2014) for Europe and North America.

7 Concluding Remarks
We have shown how height system unification can be achieved when ellipsoidal heights from
GNSS (or altimetry), orthometric heights from geodetic (or oceanographic) leveling, and geoid
heights from global geopotential models and gravity anomalies are available at each fundamental
point, and/or other stations, in each of the existing national or regional datums. Given such
observations, we also presented the functional and stochastic models used in a linear least-squares
adjustment for the determination of the offsets between existing geodetic vertical datums. The
least-squares solution provides the values of the height offsets ıNj , j D 1; 2; : : :; J , between the
various datum zones and the one adopted as the reference zone (j D 0), and the height offset No of
the reference zone geoid with respect to the reference ellipsoid. Care must be taken so that all data
are referred to a common observation epoch, reference system and reference ellipsoid, height type,
and permanent tide system, and are free from blunders and systematic effects, including temporal
changes such as those induced by postglacial rebound, geodynamics, sea level change, etc.
We explained in detail how the required geoid undulations can be obtained as the solution of
the geodetic boundary value problem for the case of multiple datums. We showed that the use
of biased, i.e., referring to individual datums, gravity anomalies introduces a direct bias on the
geoid and an indirect effect on the disturbing potential. The computation of the later is not easy
but fortunately becomes negligible when geoid undulations are computed by the remove-restore
technique using a satellite-only geopotential model with maximum degree and order of at least 70
and appropriately modified kernel in Stokes’s integral. Such geopotential models are now plentiful
and quite accurate due to the gravity field improvements brought by the dedicated gravity satellite
missions of GRACE and GOCE.
Throughout the paper, appropriate references have been given, which discuss the theory,
methodological developments and numerical studies of regional and continental vertical datum
unification. Global vertical datum unification has not been attempted yet, but this work is becoming
more relevant now that countries in Europe, Oceania, and North America have started abandoning
traditional leveling networks and adopting geoid (or quasi-geoid) based vertical datums realizable
through a geoid surface and GNSS positioning. In North America, for example, Canada replaced in
November 2013 its archaic leveling and tide gauge based vertical datum of 1928 (CGVD1928) by
a new, geoid-based vertical datum, the Canadian Geodetic Vertical Datum 2013 (CGVD2013),
as part of its Height Reference System Modernization; detailed information can be found at

Page 14 of 16
Handbook of Geomathematics
DOI 10.1007/978-3-642-27793-1_83-1
© Springer-Verlag Berlin Heidelberg 2014

http://www.nrcan.gc.ca/earth-sciences/geomatics/geodetic-reference-systems/9054. The National


Geodetic Survey of the USA has announced that they plan to do the same by 2022, and other
countries will certainly follow soon. The global unification of all of these datums is therefore
a necessity that will be realized in the near future through international collaboration between
nations.

Acknowledgments The financial supports provided by the Natural Sciences and Engineering
Research Council (NSERC) of Canada and by the European Space Agency (ESA) for the project
STSE GOCE+: Height System Unification with GOCE are gratefully acknowledged.

References
Amjadiparvar B, Sideris MG, Rangelova E (2013) North American height datums and their offsets:
evaluation of the GOCE-based global geopotential models in Canada and USA. J Appl Geodesy
7(3):191–203
Amos MJ, Feathersone WE (2009) Unification of New Zealand’s local vertical datums: iterative
gravimetric quasi-geoid computations. J Geodesy 83:57–68. doi:10.1007/s00190-008-0232-y
Argus DE, Peltier WR (2010) Constraining models of postglacial rebound using space geodesy: a
detailed assessment of model ICE-5G (VM2) and its relatives. Geophys J Int 181:697–723
Bursa M et al (2008) Mean Earth’s equipotential surface from Topex/Poseidon altimetry. Stud
Geophys et Geod 42:459–466
Cartwright DE, Crease J (1963) A Comparison of the geodetic-reference levels of England and
France by means of the sea surface. Proc R Soc Lond Ser A Math Phys Sci 273(1355):558–580
Colombo O (1980) A world vertical network. Report 296, Department of Geodetic Science and
Surveying, Ohio State University, Columbus
Featherstone WE, Filmer MS (2012) The north-south tilt in the Australian Height Datum
is explained by the ocean’s mean dynamic topography. J Geophys Res 117:C08035.
doi:10.1029/2012JC007974
Fischer I (1977) Mean sea level and the marine geoid – an analysis of concepts. Mar Geodesy 1:37.
doi:10.1080/01490417709387950
Gatti A, Reguzzoni M, Venuti G (2013) The height datum problem and the role of satellite gravity
models. J Geodesy 87:15–22. doi:10.1007/s00190-012-0574-3
Gerlach C, Fecher F (2012) Approximations of the GOCE error variance-covariance matrix for
least-squares estimation of height datum offsets. J Geod Sci 2(4):247–256
Gerlach C, Rummel R (2013) Global height system unification with GOCE: a simulation study on
the indirect bias term in the GBVP approach. J Geodesy 87:57–67
Heck B (2005) Problems in the definition of vertical reference frames. In: Proceedings of V Hotine
Marussi Symposium on Mathematical Geodesy, Matera. International Association of Geodesy
symposia, vol 127. Springer, pp 164–173
Heck B, Rummel R (1990) Strategies for solving the vertical datum problem using terrestrial and
satellite geodetic data. In: Proceedings of Sea Surface Topography and the Geoid, Edinburgh.
International Association of Geodesy symposia, vol 104. Springer, New York, pp 116–128
Hofmann-Wellenhof B, Moritz H (2006) Physical geodesy, 2nd edn. Springer, Wien/New York
Jekeli C (2000) Heights, the geopotential and vertical datums. OSU report no. 459. Depart of Civil
and Environmental Engineering and Geodetic Science, Ohio State University, Columbus

Page 15 of 16
Handbook of Geomathematics
DOI 10.1007/978-3-642-27793-1_83-1
© Springer-Verlag Berlin Heidelberg 2014

Moritz H (2013) Classical physical geodesy. In: Freeden W, Nashed MZ, Sonar T (eds) Handbook
of Geomathematics. Springer, Berlin/Heidelberg
Pail R et al (2011) First GOCE gravity field models derived by three different approaches.
J Geodesy 85(11):819–843
Pavlis NK, Holmes SA, Kenyon SC, Factor JK (2012) The development and evalua-
tion of the Earth Gravitational Model 2008 (EGM2008). J Geophys Res 117:B04406.
doi:10.1029/2011JB008916
Pugh DT (1987) Tides, surges, and mean sea level. Wiley, Chichester
Rangelova E, Sideris MG, Amjadiparvar P, Hayden T (2014) Height datum unification by means of
the GBVP approach using tide gauges. In: Proceedings of the VIII Hotine-Marussi Symposium
on Mathematical Geodesy, Rome. International Association of Geodesy Symposia, vol 142.
Springer (in press)
Rangelova E, van der Wal W, Sideris MG (2012) How significant is the dynamic component of the
North American vertical datum? J Geod Sci 2(4):281–289. doi:10.2478/v10156-012-0005-7
Rapp RH, Balasubramania N (1992) A conceptual formulation of a world height system. Report
421, Department of Geodetic Science and Surveying, Ohio State University, Columbus
Rummel R, Gruber T, Sideris MG, Rangelova E, Woodworth P, Hughes C, Ihde J, Liebsch G,
Rülke A, Schafer U (2014) STSE–GOCE+: height system unification with GOCE – summary
and final results. ESA Study Contract Report No. GO-HSU-PR-0021. Available online at: http://
www.goceplushsu.eu/gpweb/gc-cont.php?p=65
Rummel R, Ilk KH (1995) Height datum connection – the ocean part. Allg Vermessungs 8–9:
321–330
Rummel R, Teunissen P (1988) Height datum definition, height datum connection and the role of
the geodetic boundary value problem. Bull Géod 62:477–498
Sanchez L (2007) Definition and realization of the SIRGAS vertical reference system within
a globally unified height system. In: Proceedings of Dynamic Planet, Cairns. International
Association of Geodesy symposia, vol 130. Springer, pp 638–645
Sanchez L (2008) Approach for the establishment of a global vertical reference level. In:
Proceedings of VI Hotine-Marussi Symposium on Theoretical and Computational Geodesy,
Wuhan. International Association of Geodesy symposia, vol 132. Springer, pp 119–125
Sansò F, Sideris MG (eds) (2013) Geoid determination – theory and methods. Lecture Notes in
Earth-Sciences. Springer, Berlin/Heidelberg, 734pp
Sansò F, Usai S (1995) Height datum and local geodetic datum in the theory of geodetic boundary
value problems. Allg Vermussungsnachreichten, Wichmann, Heft 8–9:343–385
Sansò F, Venuti G (2002) The height datum/geodetic datum problem. Geophys J Int 149:768–775
Sideris MG, Fotopoulos G (eds) (2012) Special issue on regional and global geoid-based vertical
datums. J Geod Sci 2(4):246–376
Tenzer R, Vatrt V, Gan L, Abdalla A, Dayoub N (2011) Combined approach for the unification of
leveling networks in New Zealand. J Geod Sci 1(4):324–332. doi:10.2478/v10156-011-0012-0
Woodworth PL, Hughes CW, Bingham RJ, Gruber T (2012) Towards worldwide height system
unification using ocean information. J Geod Sci 2(4):302–318
Xu P (1992) A quality investigation of global vertical datum connection. Geophys J Int 110:
361–370. doi:10.1111/j.1365-246X.1992.tb00880.x
Xu P, Rummel R (1991) A quality investigation of global vertical datum connection. New Series,
vol 34. Netherlands Geod Commission, Delft

Page 16 of 16

View publication stats

You might also like