Computers, Environment and Urban Systems 40 (2013) 14–23

Contents lists available at SciVerse ScienceDirect

Computers, Environment and Urban Systems

journal homepage: www.elsevier.com/locate/compenvurbsys

Reprint of: Requirements of 3D cadastres for height systems q

Gerhard Navratil ⇑, Eva-Maria Unger
Vienna University of Technology, Institute for Geoinformation and Cartography, Gusshausstr. 27-29, A-1040 Vienna, Austria

a r t i c l e i n f o a b s t r a c t

Article history: Three-dimensional (3D) cadastral systems can only be built if there are suitable height reference systems.
Available online 20 May 2013 However, the earth is neither homogeneous nor flat, and the plumb lines are curved and not parallel.
Thus, the definition of height and the implementation of geodetic height systems are complex. 3D cadas-
Keywords: tres must connect to the existing national height reference. In addition, cadastral systems are designed to
Height definition persist for centuries. Over such long periods, the earth changes, from plate tectonics, erosion, human
Height system intervention, etc. Changes in the technology of measurement equipment can also occur, allowing differ-
Quality requirements
ent definitions. These considerations are important in the design of a 3D cadastre. At minimum, the
Cadastre
height system used for the cadastre must be well-defined, to enable adjustments from potential changes
in the height system or even the actual point heights.
This study examines existing height systems and the determination of height in the context of cadastral
tasks. Accuracy requirements for height in a 3D cadastre are analyzed using typical examples. The
selected height system must support these requirements. This study also develops the questions that
must be answered and highlights the problems that can emerge in some of the solutions. There is no
height reference system that fits all needs, and each solution has advantages and disadvantages. Different
systems may be optimal for different countries. It may also be beneficial to allow different geometrical
qualities for different parts of a country.
Ó 2013 Elsevier Ltd. All rights reserved.

1. Introduction which are assumed to be representations of the vertical axis, are

In discussions regarding three-dimensional (3D) systems, com-
puter scientists and technicians typically assume the existence of
suitable height reference systems. van Oosterom et al., for example,
talk about 3D objects without a clear identification of the three spa-
tial components (van Oosterom et al., 2005). This approach works
for modeling purposes but may lead to problems during implemen-
tation. Other assumptions usually not questioned in the literature
on 3D cadastres are the following: height is represented by a single
number, the direction of the vertical axis is independent of location,
the vertical axis can be represented by a straight line, and the meth-
od used to determine height is irrelevant. The validity of these
assumptions is crucial because a 3D cadastre is used to provide
sound legal protection. This cannot be guaranteed, for example, if
the results of the height observation depend on the method used.
Because the earth is neither homogeneous nor flat, the plumb lines,
DOI of original article: http://dx.doi.org/10.1016/j.compenvurbsys.2012.11.002
q
An error resulted in this article appearing in the wrong issue. The article is
reprinted here for the reader’s convenience and for the continuity of the special
issue. For citation purposes, please use the original publication details: Computers,
Environment and Urban Systems,38, pp. 11-20.
⇑ Corresponding author. Tel.: +43 (1)58801 12712; fax: +43 (1)58801 12799.
E-mail address: navratil@geoinfo.tuwien.ac.at (G. Navratil).
neither straight nor parallel. This situation causes problems in the
definition of height and the implementation of geodetic height
systems. Different observation methods refer to different height
definitions, producing varying height values with varying accuracy.
Thus, none of the typical assumptions are completely valid. While
some assumptions may be satisfactory for local applications with
limited accuracy demands, they have a much larger effect on appli-
cations where height must be determined for large areas, e.g., entire
countries.
3D cadastres encounter the problems of other 3D systems, but
additionally, cadastral systems are designed to persist for centu-
ries. Over such long periods, the earth changes considerably, and
this fact must be incorporated into the system design. Classical
two-dimensional (2D) cadastres did not face this problem because
of the following reasons:
1. 2D cadastres use an analog representation at scales where the
effects of drawing precision and line width exceed the change
in reality.
2. 2D cadastres define the boundary line itself in reality and use
the cadastral representation as an approximation only.
However, there are also 2D cadastres with fixed coordinates.
These types of cadastres also have to address earth’s changes.

0198-9715/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.compenvurbsys.2013.04.001
 G. Navratil, E.-M. Unger / Computers, Environment and Urban Systems 40 (2013) 14–23
Neunzert defined the cadastral principles for changing bound-
aries for the system in Colorado, USA (Neunzert, 2011, p. 5), as
follows:
 If nature moves a monument, the property moves with it.
 If a human moves a monument, it is an illegal act.
However, there may not always be monuments determining the
boundary. In these cases, the represented boundary in the parcel
documentation, knowledge of the land owners, and other evidence
from reality must be adapted to reconstruct the boundaries. The
importance of the different information sources may vary with
the jurisdiction.
Today, analog representations are not applicable because a
modern cadastre utilizes the benefits of computer systems includ-
ing databases and their data models (Kaufmann & Steudler, 1998).
Depending on the level of detail used for the 3D representation, the
second approach may not be valid either. Stoter and van Oosterom
present examples where the level of detail in the cadastral repre-
sentation is so high that it cannot be seen as only an approximation
(Stoter & van Oosterom, 2006, pp. 56–59). Stoter and van Oosterom
also addressed the question of height (Stoter & van Oosterom,
2006, pp. 192–195), but they approached the problem mainly from
the modeling perspective. However, geodesists have defined vari-
ous systems for height definition: geopotential heights, dynamic
heights, orthometric heights, normal heights, and ellipsoidal
heights (Hofmann-Wellenhof & Moritz, 2005, pp. 157 ff.). Each of
them has advantages and disadvantages. In this study, the various
aspects of the height problem are investigated from both a geo-
detic and a cadastral perspective to provide a solid basis for further
research. The questions posed are the following:
1. What is state of the art in the definition of height?
2. What are the benefits and disadvantages of the various height
systems for 3D cadastres?
3. What other questions must be answered both on an interna-
tional and a national level?
This paper starts with a discussion of existing height systems,
their characteristics, and how the most commonly used height
determination systems work. These differences must be consid-
ered when selecting a height system for a 3D cadastre. Section 3
then shows the typical methods to obtain heights. Section 4 intro-
duces the difference between absolute and relative heights. Accu-
racy requirements for height in a 3D cadastre are defined in
Section 5 using prototypical examples. For example, a tunnel
below the Alps requires less precision than the vertical separation
between two apartments in a building. Section 6 connects the dis-
cussion to the data model in the ISO (International Organization of
Standardization) standards, esp. ISO 19107, Geographic Informa-
tion – Spatial Schema. Section 7 presents the conclusions and some
questions for further research.
2. Definition of height
Height is usually defined as the ‘‘vertical distance from a datum
(Wolf & Brinker, 1989, p. 109).’’ This leads to several questions:
– What measurement unit shall be used to determine the
distance?
– Is distance a geometrical property?
– Which datum shall be used as the measurement start point?
– How is vertical defined?
These questions, which at first seem trivial, can offer some
challenges. Meters and gal (0.01 m/s2) are used as measurement
15
units to determine height; the geoid, the pseudo-geoid, or the
ellipsoid can be used as a datum; and the vertical line may be
assumed to be straight or bent. These differences result from dif-
ferent approaches in the definition of height, which are discussed
in the remainder of the paper.
There are several practical requirements for height systems:
1. Height (or height differences) must be easy to determine. In
everyday experience, height seems to be similar to dimensions
such as width or diameter. Basically, it should be as simple as
possible to determine the height.
2. Heights must be well-defined and path-independent, i.e., the
height difference of a closed loop should be zero. Any other
value would contradict our logical systems because it would
be possible to increase the height of a point by repeatedly com-
puting it.
3. Correctional terms for observations must be small enough to be
ignored for small surveys. Correctional terms are typically diffi-
cult to compute and require additional information. Both
aspects lead to additional computational costs and make the
process of height determination more complicated.
4. Heights must be free of hypothesis. Any hypothesis could be fal-
sified by additional observation or experiments. This effect
could subsequently lead to contradictions between previously
determined heights and new measurements.
5. Heights must be physically meaningful. For example, the idea
that water flows naturally from a point with a specific height
to another point with the same height contradicts common
sense, so it should be avoided, at least locally.
6. Heights must be geometrically determined. Unlike other prop-
erties, geometrical properties are easy to determine, and such
observations are also inexpensive.
Rod leveling is a simple method to determine height differ-
ences. It uses a plumb line to determine the vertical direction but
contradicts the second requirement that closed loops should have
no height difference, as shown in Fig. 1. All points on any equipo-
tential surface have the same potential energy, but the equipoten-
tial surfaces are not parallel. The plumb lines between points A and
A0 and points B and B0 are perpendicular to the equipotential sur-
faces, and because the surfaces are non-parallel, these plumb lines
are curved. The lowest equipotential surface shown in Fig. 1 is at
sea level and is the geoid (defined as the average sea level). The
thick line connecting points A and B represents the surface of the
earth. Assume that the height difference between points A and B
is determined as a length along the plumb line. The length between
points A and A0 is different from the one between points B and B0 .
Rod leveling would determine the height difference in small steps
along the surface, resulting in a value between these extremes.
This could not happen if the second requirement holds.
A number of different height systems have been proposed that
fulfill the above requirements. The most commonly used systems
are the following (Hofmann-Wellenhof & Moritz, 2005, pp. 159 ff):
– geopotential height,
– dynamic height,
– orthometric height,
– normal height, and
– ellipsoidal height.
2.1. Geopotential height
The geopotential height or geopotential number, C, of a point is
the potential difference of the point’s potential energy and the
geoid’s potential energy. Because it is a potential difference, the
geopotential number, C, is path-independent. The geopotential
16 G. Navratil, E.-M. Unger / Computers, Environment and Urban Systems 40 (2013) 14–23
Fig. 1. The non-parallel nature of equipotential surfaces and its effect on plumb lines and heights.
number is typically measured either in J/kg (Joule per Kilogram),
in m2/s2, or in gal m. gal is a measure for acceleration, and its
definition is cm/s2. Although the potential difference does not
describe a distance, it is a natural criterion for heights because
on a surface with constant potential energy, objects do not start
to move if no force is applied and, in particular, water does not
flow.
The disadvantages for geopotential heights are the difficult
measurement process and the above-mentioned character that it
does not describe a distance. The conversion between gal and
meters can be done by applying the following conversion:
h½m ¼ a  C½gal  m þ b;
but unfortunately, the parameters a and b are location dependent.
The parameter a also varies with the height because gravitational
acceleration changes by approximately 0.3 mgal per meter at sea le-
vel. Because the earth’s tides produce a variation in the geopotential
height of up to 0.3 mgal (Wolf, 1940), geopotential height is theo-
retically not constant itself. The variation is caused by the attraction
of the moon and sun (compare Agnew, 2007).
Determination of geopotential heights requires a combination
of geometrical leveling and gravity measurements. The leveling
provides the distance and the gravity measurement the accelera-
tion. Therefore, continuous observations along the path would be
necessary for exact measurements. Discrete observations are suffi-
cient for practical implementation if the distance between the dis-
crete points is not too large.
2.2. Dynamic height
Dynamic heights are derived from geopotential numbers as
follows:
C local measurements. Thus, local measurements only require geo-
Hdyn ¼ ; metrical leveling to measure orthometric heights. Therefore, ortho-
c0
where C is the geopotential number and c0 is the normal gravity for
an arbitrary standard latitude. Typically the normal gravity at a 45°
latitude is used:
c45 ¼ 9:806 199 203 ms2 ¼ 980:619 920 3 gal:
The division by c0 converts the geopotential number into a
length. Because it is a scaled geopotential number, it still has phys-
ical meaning, the determination is still path-independent, and the
measurement process is the same as that for geopotential height.
The use of an arbitrary value for the scaling keeps the system free
of any hypothesis regarding the mass distribution within the earth.
However, dynamic height has no geometrical meaning. When
translating vertical distances into differences of dynamic height,
the correctional terms can become large. Hofmann-Wellenhof
and Moritz show that the correction for a height difference of
1000 m can easily be in the range of several meters (Hofmann-
Wellenhof & Moritz, 2005, p. 169).
2.3. Orthometric height
Orthometric height is defined as
C
H¼
c
where C is the geopotential number and c  is the average gravita-
tional acceleration along the plumb line between the geoid and
the observed point. The determination of c
 is
Z H
1
c ¼ cðzÞ dz:
H 0
Calculation of the orthometric height would thus require
detailed knowledge on the mass distribution in the earth along
the path to determine c(z). Because the distribution is not known,
a variety of approximations has been developed. The simplest
version is the Prey reduction (Hofmann-Wellenhof & Moritz,
2005, p. 139). It uses the arithmetic mean of the gravity c, observed
at the surface, and the gravity c0, computed at the corresponding
geoidal point, to estimate c .
Using the orthometric height has several practical advantages:
– point heights are well-defined and path-independent,
– point heights define a geometrical property, and
– corrections of the observed values are small.
The corrections are small enough that they can be neglected for
metric heights were the best option for height systems in the 20th
century for a large number of countries. However, orthometric
heights also have some weaknesses. Water may flow between
points with the same orthometric height, and orthometric heights
are often not free of hypotheses.
2.4. Normal height
The height systems described above use the geoid as a reference
surface. Normal heights use an approximation of the geoid, i.e., an
ellipsoid, as the reference surface. The basic assumption is that the
earth has a regular gravity field so that the potential, W, of the
geoid is equal to the potential, U, of the regular gravity field. The
potential difference is then
 G. Navratil, E.-M. Unger / Computers, Environment and Urban Systems 40 (2013) 14–23
Z H
W0  W ¼ C ¼ cdH :
0
c is the gravity at the ellipsoid. The normal heights, H⁄, can then be
defined as
Z C
dC
H ¼ : An approach to track this issue is the creation of a unified global
0 c
Any given point B on the surface has a potential WB and a normal
potential UB. These two potentials are different because irregulari-
ties of the earth’s gravity field are neglected in the regular gravity
field. There is a point B, however, where the normal potential U B
equals the true potential W B . The normal height HB of the point
B is then equal to the orthometric height of B on the ellipsoid using
the regular gravity field.
No hypothesis is required for this definition because a regular
gravity field is used. Other advantages are the independence of
the path and the small corrections. However, the height has no
physical meaning and it is not geometrically determined. Another
problem is that the height of a point is determined by a length that
does not end at this point. To eliminate this problem, heights are
applied downward from the terrain. This strategy leads to the
quasigeoid, which coincides with the geoid at the oceans but devi-
ates from it everywhere else. The vertical distances between the
geoid and the quasigeoid, i.e., the anomalies, can be up to
40–50 cm in mountains, such as the Alps (Bretterbauer & Weber,
2003, p. 37).
2.5. Ellipsoidal height
Ellipsoidal heights have a purely geometric definition. Height is
the distance between the point and the ellipsoid measured along
the surface normal of the ellipsoid. Obviously, ellipsoidal heights
have no physical meaning. Their advantages are the path-indepen-
dent determination and the absence of any hypothesis. However,
deriving heights based on known points and local observations of
height differences can require large corrections because locally
observed height differences typically refer to the direction of the
plumb line and not to the direction of the ellipsoid’s surface nor-
mal. Ellipsoidal heights are produced by global navigation satellite
systems (GNSS), such as GPS, GLONASS, or GALILEO. Thus, they are
easy to determine with common GNSS-receivers.
2.6. Summary and conclusions
Table 1 summarizes the practical requirements for heights de-
fined in Chapter 2.
This overview shows that the definition of height is neither sim-
ple nor straightforward. There is no height definition that produces
physically meaningful heights that are easy to measure and that do
not require a correction of the observations.
Conversion between the different systems is not trivial either.
The different reference surfaces as well as the curvature of the
plumb lines must be addressed in most cases. The vertical differ-
ence between geoid and ellipsoid, i.e., the geoid undulation, can
Table 1
Fulfillment of the practical requirements by the different height systems.
Geopotential
Easy to determine No
Well-defined and path-independent Yes
Small correctional terms – No
Hypothesis-free Yes
Physically meaningful Yes
Geometrically determined No
17
be nearly 200 m when using a reference ellipsoid. This issue must
be addressed, for example, when deriving orthometric heights
from ellipsoidal heights. The determination of the geoid undulation
in Croatia, for example, produces an accuracy of approximately
10 cm (Grgić et al., 2009). Thus, a combination of GNSS height
observations with terrestrial observations creates uncertainties.
height reference, which is required for the Integrated Global Geo-
detic Observation System (IGGOS). Such a system provides all of
the data necessary to convert between different height systems
(Ihde & Sánchez, 2005). A proposal for the Association of Southeast
Asian Nations (ASEAN) assesses an accuracy of 13 cm for distances
of 500 km and assumes a higher accuracy for shorter distances
(Lan, 2009). New Zealand implemented such a system to overcome
the limitations of the prior existing 13 local height systems (Amos,
2009). The standard deviation of the offset to the old systems has
been assessed as less than 10 cm (with one exception, where the
standard deviation is 15 cm).
3. Determination of height
The determination of height uses measurement equipment that
can observe a specific quantity, e.g., length or gravity. The geomet-
rical determination of height is typically the easiest. A horizontal
plane, defined by a level, is used as an approximation for points
of equal height, and then vertical deviations to that plane are
observed. This is the principle of geometrical leveling as shown
in Fig. 2a. The drawback of geometric leveling is that it does not
strictly fit to any of the height systems defined in Section 2 because
the horizontal plane is only an approximation of the reference
surface. The longer the observation distance, the more error is
introduced by the approximation. Thus, very accurate height deter-
mination is rarely possible at observation distances greater than
10 m. For large leveling networks, gravity must be measured along
the leveling routes to further reduce the effects of approximation.
Still, geometrical leveling is used to determine local height differ-
ences for orthometric and normal height systems because the
accuracy of the approximation is sufficient for most applications.
A similar method that strictly fits the physical definition of height
is hydrostatic leveling, where the equipotential surface is not
approximated by a plane but visualized by a hose filled with a li-
quid, typically water (Fig. 2b). Hydrostatic leveling is mainly used
for high precision deformation monitoring (compare Meier et al.,
2011), and it has been applied in long-distance measurements
(Waalenwijn, 1964).
The determination of ellipsoidal heights must refer to the sur-
face normal of a specified ellipsoid. Measuring the surface normal
in reality is impossible except for very specific points where the
surface normal of the ellipsoid and the plumb line coincide. Other-
wise, the direction can only be determined mathematically. GNSS
is a typical method to determine ellipsoidal height. GNSS measures
the distance between the unknown position of the receiver and the
known satellite positions and computes the position of the receiver
in a geocentric coordinate system. The coordinates can then be
used to determine ellipsoidal latitude, longitude, and height if
Dynamic Orthometric Normal Ellipsoidal
Yes Yes Yes Yes
Yes Yes Yes Yes
Yes Yes Yes
Yes No Yes Yes
Yes No No No
No Yes No Yes
18 G. Navratil, E.-M. Unger / Computers, Environment and Urban Systems 40 (2013) 14–23
Fig. 2. Leveling methods.
the relation between the ellipsoid and the geocentric coordinate
system are known. Direct observation of the ellipsoidal height dif-
ference is not possible because the direction of the vertical is not
visible. However, ellipsoidal height differences can be computed
as the difference of ellipsoidal heights. Geometrical leveling pro-
vides an approximation of ellipsoidal height differences if the
deflection of the vertical is known in the area of interest and if
the area is not too large.
The practical advantage of geometrical leveling is that it is
applicable everywhere and it may be one of the cheapest methods
to measure height differences in terms of time, equipment, and
training. Geometrical leveling is possible both inside and outside
of buildings or tunnels and can provide height information where
satellite-based measurement methods fail. In addition, height
determination is possible with high precision if required, which
continues to be a problem for GNSS. However, satellite-based or
airborne methods have an advantage if the heights of many points
in large areas must be determined.
Additionally, the geoid is not stable and is changing from natu-
ral processes. Natural changes to the geoid have been evaluated by
Jacob et al. (2012) and his team in the study ‘‘Estimating geoid
height change in North America: past, present and future’’.
This study, although focused on North America, can show the
influence and impacts of natural processes on the geoid and the re-
lated defined heights. The main factors for a changing geoid (the
GRAV-D gravimetric geoid model over the United States was used)
have been determined:
 Errors in a glacial isostatic adjustment model induce error on
the order of 10 mm in the geoid height level after approxi-
mately 20 years.
 Geoid height change over the ice mass loss regions of North
America may lead to changes on the order of 10 mm in height
in under a decade.
 The computed sea level rise and its effect on the geoid are found
to be negligible.
 Co-seismic gravitational changes due to earthquakes may lead
to geoid changes on the order of 10 mm, dependent on the mag-
nitude of the earthquake.
 Volcanic processes were found to be significant.
Because the earth is continually changing, reference systems
have a temporal component in addition to the spatial component.
For example, the European Terrestrial Reference Frame (ETRF) is
available in different versions corresponding to different years. Be-
cause the value of measurements may change with time, they must
have a time stamp. Changes of the earth can then be tracked by
comparing measurements made at different times. The effects of
these changes can then be eliminated, and improved information
on local objects can be provided.
4. Relative vs. absolute height definition
Height can be defined either in absolute terms or relative to the
height of another point. Geopotential numbers can be seen as
absolute heights, although they vary with the tides. Strictly speak-
ing, all other height systems discussed in Section 2 are relative sys-
tems because their height is related to a specific surface, the geoid
or an ellipsoid. However, this is not what is typically meant by rel-
ative height. Usually, statements such as ‘‘5 m below the ground’’
are seen as relative definitions. These values are much easier to
measure than absolute values because, in most cases, they can be
measured directly. If the distance becomes too large, relative
heights may be difficult to determine, e.g., the depth of the
Gotthard Base Tunnel below the ground at a given point. The
answer not only varies with the position in the tunnel, but it is also
difficult to measure. National height systems typically strive to
provide absolute heights. However, in times of international
cooperation and trans-national height systems, the term ‘‘national
system,’’ although still in use, may be misleading.
Stoter and van Oosterom propose the use of absolute height
coordinates (Stoter & van Oosterom 2006, p. 192f). Absolute height
coordinates can be challenging, however, because the earth is not
static. Scandinavia, e.g., rises at a rate of up to 1 cm/year (Lidberg,
Johansson, Scherneck, & Davis, 2007). Height coordinates should
therefore change with the same rate. However, they typically have
fixed values. The system only works in practice because the coor-
dinates are derived locally using reference points that are also af-
fected by the movement. Thus, although the absolute coordinates
provide incorrect values, the difference between the coordinates
is correct due to the high spatial autocorrelation of the errors. Re-
peated height determination in Austria proved the vertical move-
ment of control points. Imrek presented results from an
experiment near Lake Attersee in Upper Austria. All inspected
points showed vertical movements. However, the amount of move-
ment varied between 5 cm and 25 cm for a time period between 10
and 50 years. This result suggests a movement rate of up to 5 mm/
year for this area. Based on this and other experiments, Imrek
assumes that 10% of all control points in Austria are subject to
vertical movements (Imrek, 2012). The problem with these move-
ments is that the vertical movements are irregular. Horizontal
movements are mainly caused by plate tectonics, and thus, large
areas move in a unique way. Vertical movement is structured
much finer, and thus, discrepancies are more difficult to model.
In everyday life, local height systems are more common.
Building permits are usually restricting the height of buildings in
relation to the surface. This may be done by absolute heights if
the permit contains a detailed architectural design, but the heights
are then determined by adding the height values of the design to
 G. Navratil, E.-M. Unger / Computers, Environment and Urban Systems 40 (2013) 14–23
the known absolute heights. Another example for the use of local
height systems is the level of groundwater, which can be expressed
as ‘‘a depth of 3 m.’’ In many cases, relative measures may not use
length units to specify the height. ‘‘My apartment is on the second
floor’’ is a relative height definition that is often used and easy to
check by observation. However, in terms of meters, it may corre-
spond (in Vienna) with a value anywhere between 2 and 4 m. It
should be noted, though, that even this simple method contains
pitfalls. Counting may start with zero (ground floor) or one (first
floor), and legal regulations may lead to a situation such as that
in Vienna, where, between the ground floor and the first floor,
many old houses have an intermediate level, a mezzanine, or both.
This issue will be relevant for the ownership of the apartment in
the example in Section 5.4.
A problem for relative height definition with respect to the
surface of the earth is that the surface may change. The reason
for the height change may be a natural phenomenon, e.g., a land-
slide. In extreme cases, e.g., if parts of mountains slide downward,
this may result in height differences of several meters. Fig. 3 is a
schematic representation of the problem. A volume dV slides from
a higher to a lower position, thereby changing the terrain surface.
The vertical difference between the old and new terrain surface is a
height change that must be applied to relative height values in the
affected area. A height change can also occur because of human ac-
tions, e.g., construction. Road refurbishment, for example, may
change the vertical position of the road either to minimize vertical
movement or because a new base layer had to be created. This
problem occurs for objects that are close to the surface or need
to be defined precisely, as will be shown in Sections 5.2–5.4.
5. Accuracy requirements for height in 3d cadastres
A modern 3D cadastre must fulfill the requirements of e-gov-
ernment. Digital cadastral maps are typically part of an e-govern-
ment infrastructure (compare Austrian Federal Chancellery, 2011,
p. 85). To provide a working spatial reference, the basis of the
cadastral maps must be the official reference system. This is typi-
cally a plane coordinate system for classical 2D cadastres. A 3D
cadastre also requires a vertical reference system. The use of differ-
ent reference systems within one jurisdiction may have negative
side-effects, e.g., if the same physical object has a different vertical
extent based on the reference system used. This can lead to legal
problems because it can be seen as a contradiction to the principle
of equal treatment. Thus, 3D cadastres have to utilize a common
national height reference system. An internationally defined
system could be used, but its use may lead to inconsistent data sets
when incorporating old data sets that are based on a national
height definition. The conversion of the old data sets to any inter-
nationally defined system would have only limited accuracy. The
Austrian Federal Office of Metrology and Surveying (BEV) provides
Fig. 3. Height change caused by landslides.
19
a grid-based solution for the conversion. The standard deviation of
the conversion is 2 cm up to medium heights. In the Alpine regions,
they expect the standard deviation to be higher (Briese, Höggerl,
Imrek, Otter, & Ruess, 2012). The Austrian highest administrative
court classified a horizontal deviation of 4 cm from a required
4 m clearance as significant (VwGH, 1983). Assuming that these
are the lowest reasonable requirements, then the solution pre-
sented by the BEV is sufficient if the conversion is only done once.
Significant effort was required to produce this solution. The geopo-
tential height of more than 5,000 points was known from the Euro-
pean Leveling Network 95/98 (UELN95/98). This point network
was complemented by over 42,000 national height points. In addi-
tion, approximately 50,000 measured values for gravitational
acceleration were used to interpolate the grid. Because Austria
has a size of 83,000 km2, there was a gravity measurement and a
control point from an adjusted leveling network every 1.5 km2
(Briese et al., 2012; Höggerl, 2012). One problem with the grid is
the adaptation to the vertical movement of the earth. Using the
movement of 0.5 cm/year and demanding an accuracy of 5 cm
leads to an update cycle of 10 years for the height grid. Thus, there
are multiple grids that must be used for data measured at different
times, complicating data collection considerably.
The height definition in the cadastre must, at a minimum, sat-
isfy the accuracy requirements of the cadastre. Ever since the dis-
cussion of 3D cadastres emerged at the end of the 20th century, the
accuracy requirements of the cadastre are still under discussion. To
date, there is no practical experience from lawsuits; thus, there
may be opinions expressed by lawyers in the future, but there
are no valid legal conclusions to date.
The accuracy requirements for plane coordinates can be used as
the basis for the 3D accuracy requirements. Coordinates in Austria
are registered with centimeter precision. This corresponds to the
geometrical accuracy of average construction work, such as placing
walls. Precision for boundaries of farmland can be less accurate be-
cause boundaries of farmland need to be known with a precision of
approximately 20 cm (less than the width of a plow blade). For
land owners, the shape of the parcel is more important than the
absolute geographical position. The shape determines possible
uses, e.g., if a specific building can be constructed. This perspective
is used, for example, if a court decides on the compliance with
clearances to the boundary defined by law (Navratil, 2008). The
conclusions on the accuracy of plane cadastral coordinates are
therefore as follows:
– Accuracy requirements for a cadastre are not uniform through-
out the country; and
– Local accuracy is more important than global accuracy.
Both conclusions may also be true for the vertical dimension in
3D cadastres. The four examples that follow illustrate the varying
needs for accuracy. The examples describe artificial objects that
are shown in the 3D cadastre. The examples are differentiated by
their position in relation to the surface and their dependence on
other 3D objects. The first two examples are below ground, the
third one is above ground, and the last example (ownership of an
apartment) can be both but typically requires structural support
by other objects.
5.1. Underground object deep below the surface
The Gotthard Base Tunnel mentioned in Section 4 is an example
of an underground object that is deep below the surface. The ver-
tical distance between the tunnel and the surface is large enough
that the existence of the tunnel does not directly influence the land
use above. The only perceptible effect could be vibration caused by
the movement of the trains. Registration of the tunnel is advisable
20 G. Navratil, E.-M. Unger / Computers, Environment and Urban Systems 40 (2013) 14–23
to avoid subsequent legal trouble. The accuracy of the height def-
inition required for this tunnel is determined by the purpose of
the registration. It is only necessary to inform the owner of the sur-
face land that there is a tunnel. However, the accuracy demands
may increase if the land owner creates a construction with an
extensive foundation. In general, accuracy demands are low en-
ough that changes of the surface itself do not create a problem
for spatial reference. Even landslides with vertical effects of several
meters would not invalidate the information. The registration
quality should not be confused with the quality demands for crea-
tion or maintenance of the tunnel, however.
5.2. Underground object close to the surface
Underground objects such as tunnels cause more problems if
they are close to the surface. Hackl (2007), pp. 34–26) presented
an example of a road tunnel that is only a few meters below the sur-
face. The area above the tunnel is used as a playground for the adja-
cent kindergarten. Light construction work, e.g., to build a climbing
frame, is possible, but there are limitations to the depth allowed for
the foundation of these objects. Information on the depth of the
tunnel is essential to avoid damage to the tunnel structure.
Similar tunnels are built below cities, e.g., for the metro. Land
above the tunnel is typically required for buildings. Fig. 4 shows
an example of a metro tunnel below a district heating plant. In this
case, the infrastructure already existed and the building was
placed on top of the metro. Vertical clearance distance must be
considered in the construction process, including a tolerance to
cover subsidence and construction accuracy. Information on final
clearance must be more precise in the planning and implementa-
tion phase to avoid problems after finishing the project. If a tunnel
is built below existing buildings, then detailed knowledge on the
static behavior of the buildings is required. Vienna, for example,
has a large number of old buildings with up to three basement lev-
els. To avoid damage to the buildings, vertical clearances for any
construction work are required. These clearances may vary with
the ground structure and stability. An example of such a construc-
tion below an already existing building is the extension of the
Fig. 4. Metro tunnel situated below a district heating plant in Vienna.
metro below the Ringturm in Vienna. During the construction
phase, the front part of the 90-m-high building subsided 2 cm.
The back side of the building remained at the same height, and
thus, the building leaned slightly. The effect is a vertical misalign-
ment of approximately 4 cm at the top of the building.
In a built-up environment, vertical surface changes for relative
heights are problematic. Cities are constantly renewed, and the
absolute vertical position of the surface is thus changing. Even if
only a new coat of tarmac is put on a road, the vertical change
can be up to a few centimeters. Thus, use of relative heights requires
thorough knowledge and inclusion of the update processes in cities.
5.3. Construction above a public space
Similar problems occur if constructions extend into the space
above public areas such as streets. In such cases, clearances must
be met as well. Roads, for example, may have a height limit, and
it must be ensured that vehicles that comply with this limit can
use the road without any problems. Absolute height is irrelevant
as long as the relative clearance is met. Compared to underground
constructions, checking the compliance of constructions with ver-
tical clearances is easier. Legal decision making does not consider
measurement quality (Twaroch, 2005). Thus, the accuracy require-
ments must be high enough to avoid lawsuits from the relative
viewpoint.
5.4. Ownership of an apartment
Ownership of an apartment in 3D cadastres allows separation of
spaces owned by different persons within a building. Usually, this
separation is vertical and the border must lie within the ceiling.
This can be guaranteed by either high accuracy or a definition that
the surface described in the cadastre lies within the ceiling. The lat-
ter definition adopts the principle of general boundaries (compare
Land Registration Act, 2002, Section 60).
An easy-to-provide vertical reference would be counting the
floors, e.g., ground floor owned by person X, first floor owned by
person Y, etc. However, this may be difficult if the separation is
not strict, e.g., in the case of an apartment building with three
apartments, all of which have some section of the first floor, shown
by Stoter and van Oosterom (2006, p. 41). Other problems occur if
buildings are placed along hillsides. It may then not be clear where
to start counting the floors. This problem may be solved by strict
rules on counting the floors, but such rules do not exist in all coun-
tries. However, the problem that the first floor in the Stoter and van
Oosterom example consists of three apartments, although only one
is limited to the first floor, is still unresolved.
Absolute heights may also be difficult in this situation. If the
building subsides, absolute heights may lead to a situation where
the boundary moves out of the ceiling. The legal consequences of
such an incident are not clear. In countries with a coordinate-based
boundary definition, it may be necessary to correct the heights.
This would correspond to similar situations in 2D where buildings
glide down hills. The problem may be less significant in the case of
general boundaries because boundary markers typically move with
the surroundings.
6. Implementation issues
The examples in Section 5 show that the accuracy demands
vary dramatically from a few centimeters to a dozen meters. This
issue must be addressed by any 3D cadastre because otherwise,
the highest demands must be used to stipulate accuracy require-
ments. This would lead to unnecessarily expensive surveys. From
a mathematical or computing perspective, a system with varying
 G. Navratil, E.-M. Unger / Computers, Environment and Urban Systems 40 (2013) 14–23
accuracy seems to be strange. However, this strategy has already
been adopted in the 2D-cadastres established in the 19th century.
The Austrian cadastre, for example, was intended solely as a tool
for fair land taxation, and the method used to guarantee correct-
ness of the data was mapping the data. All parcels were drawn
on map sheets to guarantee completeness of the data and to make
data manipulations (e.g., changing the size of the land parcel)
difficult. The amount of land tax was based on the profit that
everyone could achieve. Thus, there was a strict separation be-
tween productive land, especially farmland, and unproductive land
such as mountain tops and residential areas. Although unproduc-
tive land was also included in the cadastral maps, quality require-
ments were lower, resulting in lower effort and lower data
accuracy. This is especially true for alpine areas where low quality
demands meet large land parcels. As a result, a mapping scale of
1:5760 was applied, whereas most of the productive land was
mapped in a scale of 1:2880.
The same principle is also used for topographical mapping.
Areas of high importance, e.g., city centers, are mapped in larger
scales than less important areas. The same is true for nautical
maps, where the most accurate maps are available for ports. The
reason in this case is that the water depth has a direct impact on
travel safety. When less depth is available, it needs to be known
more accurately. Because large ships are difficult to maneuver,
accurate course planning through narrow paths is necessary,
requiring accurate data, i.e., of high spatial resolution.
Currently, cadastral data are stored in databases according to
the concepts of cadastre 2014 (Kaufmann & Steudler, 1998).
Therefore, data modeling is increasing in importance. ISO provides
standards required for data modeling. The data type DirectPosition,
defined within ISO 19107 Geographic Information – Spatial
Schema, can be used for implementation.
Fig. 5 shows the current data model.
(1) The data type DirectPosition contains two attributes, coordi-
nate and dimension. The coordinate attribute is a sequence
of numbers that are the coordinates of the position within
Fig. 5. Excerpt from the ISO standard on spatial schema of geographic information. Source: ISO, 2005, adapted by the authors
21
the specified reference system. The dimension attribute is
the length of the coordinate sequence, determined by the
reference system.
(2) The association role, coordinateReferenceSystem, defines the
coordinate system used for the coordinate. The type SC_CRS
is defined in ISO 19111 Geographic Information – Spatial,
referenced by coordinates.
(3) The DataType GM_PointRef is used to reference an existing
point.
(4) This is an instantiation of the template class Reference
<GM_Point>.
(5) The data type GM_Position is most relevant for storing dif-
ferent types of heights.
(6) This union type consists of either a set of coordinates
(DirectPosition) or an indirect coordinate definition (a refer-
ence to a GM_Point). An indirect reference can be used to
determine coordinates, e.g., by adding an offset to the coor-
dinates of the reference point. Using this data type allows an
option to provide a position directly as a coordinate or indi-
rectly as a reference to a GM_Point.
The data model can handle both absolute and relative coordi-
nates. However, the data model may have to be expanded for 3D
cadastres. Due to different accuracies, additional information on
the actual accuracy of the point height should be stored within
the model. This can be done, for example, by defining accuracy
classes and specifying the class for each point height.
For example, the method of defining the point (e.g., Theodolite,
GNSS) should be included. Additionally, the defined accuracy of the
method should be known. The date of the measurement may also
have to be included. With all of these additional attributes, it will
be easier to react to changes from natural or human actions.
The examples in Section 5 showed that relative height
information has a higher practical importance than absolute height
information. This may, however, cause specific problems: Assume
a metro is operating directly underneath a street. This is not
unusual because of the following:
22 G. Navratil, E.-M. Unger / Computers, Environment and Urban Systems 40 (2013) 14–23
Fig. 6. Example of the different impressions from real 3D and relative height
differences.
– There is less physical pressure on the tunnel compared to the
situation where a building is located above the tunnel;
– Construction is easier because it may even be possible to create
the metro with the trench method; and
– The metro and road network may be owned by the same
organization, thereby simplifying the legal decision making.
However, roads may have larger inclinations than metro lines.
Thus, the vertical distance between the metro line and road surface
will vary along the road on hilly ground. Absolute vertical coordi-
nates would show that the track of the metro in 3D is smoother
than the road. Relative heights, however, would show a different
behavior (compare Fig. 6). The example also shows that a single
height for the object is not sufficient because the legal object
may have a large horizontal extent even if it is always cut on the
edges of the 2-dimensional parcels.
A practical problem is obtaining the height of the surface and
the legal objects that shall be registered. Terrain height is currently
typically obtained by Airborne Laser Scanning, ALS (Kraus & Pfeifer,
2001), and height determination of objects is often performed with
GNSS receivers. Both types of systems create ellipsoidal heights.
National height reference frames, however, are frequently based
on more physically defined height systems, and the conversion
between these systems is not simple. This may lead to practical
problems when implementing 3D cadastres. Tunnels cannot
usually be measured with these methods, although some tunnels
are built with the trenching method and then GNSS and ALS can
be used. This is only a problem for heavily constructed areas.
7. Discussion and conclusions
This study presents the currently existing height systems (Sec-
tion 2). Current trends in geodetic research provide more precise
and better integrated systems. Improved accuracy implies that
the temporal component of the reference system must be adopted
in the application domain. A boundary in a 3D cadastre can consist
of parts that are determined at different times. A temporal compo-
nent could result in gaps in the boundary. The benefits and disad-
vantages of the systems have been discussed briefly from a user’s
perspective. Land owners are usually laypersons, and checking
cadastral information on correctness should be possible for them.
The analyses lead to a number of questions that require further
analysis.
There is no height reference system that fits all needs. Small and
flat countries may be able to use any reference system, whereas
other countries must take existing data into account and use the
height system suitable for them. Many practical questions require
relative information only, while national systems usually provide
absolute information. Relative information is typically easier to ob-
tain and interpret by laypersons. The problem encountered when
converting relative to absolute information is tracking the height
changes of the surface. The difference between absolute heights
provides the relative height information but requires current
height information on the surface itself. It may be necessary to
use different height systems in parallel, storing values in different
systems with one cadastral object. The ISO standard on spatial
schema of geographic information is capable of managing these is-
sues, but it is necessary to address the practical problems from the
start of the 3D cadastral implementation. In practice, the existence
of different height systems will call for transformation routines to
convert data between different systems with the required accu-
racy. This may call for additional investigations on the shape of
the earth in the region. The same is true for different levels of accu-
racy, which also have to be stored in the system. All of these as-
pects can be solved but should be started in the design phase of
the system.
The discussion shows that data in a 3D cadastre will not neces-
sarily be of homogeneous accuracy. Classical 2D cadastral systems
have spatial variation in data accuracy, usually depending on the
value of the land. The measurement-based approach (Buyong,
Frank, & Kuhn, 1991) was an attempt to address this issue but
has not yet been implemented for a working cadastre. The situa-
tion worsens in a 3D cadastre because the vertical height is natu-
rally variable. The information system needs to be able to store
the accuracy parameters and to automatically determine the accu-
racy of derived values. Because correlation must be considered in
this process (compare Frank & Navratil, 2011), an accuracy-aware
information system is necessary. Analog cadastral maps contain
implicit accuracy-awareness by using a scaled representation.
The accuracy of the data collection processes matches the accuracy
required for the mapping scale. The data interpretation and extrac-
tion by the user is limited to the mapping scale. These advantages
of analog systems vanished when moving to digital methods, lead-
ing to increased quality problems.
In Section 6, an entry point for dealing with different height sys-
tems in ISO 19107 was illustrated. The discussion only shows that
the standard is capable of handling the issues encountered when
using 3D cadastres. A detailed technical analysis of implementa-
tion and accuracy is still required. Such an analysis can bridge
the gap between the available quality of vertical reference systems
and the quality demands of national regulations for 3D cadastres. A
generic discussion may be difficult because the national regula-
tions may impose varying regulations. Coordinate-based and gen-
eral boundary-based approaches have different demands and
may need to be addressed separately.
References
Agnew, D. C. (2007). Earth tides. In G. Schubert & T. Herring (Eds.), Treatise on
geophysics (pp. 163–195). Oxford: Elsevier.
Amos, M. (2009). New Zealand vertical datum 2009. New Zealand surveyor, 300,
5–16.
Austrian Federal Chancellery (2011). Administration on the net. The ABC guide of
eGovernment in Austria. <http://www.digitales.oesterreich.gv.at/publications>.
Bretterbauer, K., & Weber, R. (2003). A primer of geodesy. Geowissenschaftliche
mitteilungen (Vol. 64). Vienna, Austria: Springer. P. 50.
Briese, C., Höggerl, N., Imrek, E., Otter, J., & Ruess, D. (2012). Transformation von
GNSS-Höhen in österreichische Gebrauchshöhen mittels einer
Transformationsfläche (Höhen-Grid). Österr. Zeitschr. F. Vermessung u.
Geoinformation, 99, 267–281.
Buyong, T., Frank, A. U., & Kuhn, W. (1991). A conceptual model of measurements-
based multipurpose cadastral systems. Journal of the Urban and Regional
Information Systems Association (URISA), 3, 35–49.
Frank, A. U., & Navratil, G. (2011). Ignoring correlation leads to bone shaped
confidence regions and other couter-intuitive aspects of spatial data quality. In
C. C. Fonte, L. Goncalves, & G. Goncalves (Eds.), 7th International symposium on
spatial data quality (pp. 95–100). Coimbra, Portugal: INESC Coimbra.
 G. Navratil, E.-M. Unger / Computers, Environment and Urban Systems 40 (2013) 14–23
Grgić, I., Lučić, M., Repanić, M., Bjelotomić, O., Liker, M., Markovinović, D., et al.
. Preliminary works on creation of the new croatian geoid model HRG2009. EUREF
Symposium. Firenca, Italy: International Association of Geodesy.
Hackl, M. (2007). Dreidimensionaler Kataster – Möglichkeit, Notwendigkeit und
Fortschritt der Einführung in verschiedenen Ländern (engl. 3D Cadastre –
Possibility, Necessity, and Progress of Implementation in Different Countries).
Unpublished Diploma Thesis, Vienna University of Technology, Vienna.
Hofmann-Wellenhof, B., & Moritz, H. (2005). Practical geodesy. Vienna: Springer.
Höggerl, N. (2012). Das Europäische Vertikale Referenzsystem (EVRS) und dessen
Bezug zu den Höhen des MGI-Systems in Österreich (engl. The European
Vertical Datum (EVRS) and ist relation to the heights of the system MGI in
Austria). Presentation at the ‘‘Impulsveranstaltung: Neue Referenzsysteme’’.
<http://www.bev.gv.at/pls/portal/docs/PAGE/
BEV_PORTAL_CONTENT_ALLGEMEIN/0100_NEWS/NEWS-PDF/
HOEGGERL_KOORDINATENSYSTEME2012-10-10_EVRS07.PDF> Accessed
19.10.12.
Ihde, J., & Sánchez, L. (2005). A unified global height reference system as a basis for
IGGOS. Journal of Geodynamics, 40, 400–413.
Imrek, E. (2012). Das europäische Referenzsystem ETRS89 und dessen Bezug zum
österreichischen System der Landesvermessung MGI (engl. The European
Reference System ETRS89 and its relation to the Austrian system of land
administration MGI). Presentation at the ‘‘Impulsveranstaltung: Neue
Referenzsysteme’’. <http://www.bev.gv.at/pls/portal/docs/PAGE/BEV_PORTAL_
CONTENT_ALLGEMEIN/0100_NEWS/NEWS-PDF/
IMREK_KOORDINATENSYSTEME2012-10-10_ETRS89.PDF> Accessed 19.10.12.
ISO (2005). Geographic information – Spatial schema. EN ISO 19107:2005 (p. 174).
Jacob, T., Wahr, J., Gross, R., Swenson, S., & Geruao, A. (2012). Estimating geoid
height change in North America: Past, present and future. Journal of Geodesy, 86,
337–358.
Kaufmann, J., & Steudler, D. (1998). Cadastre 2014. FIG.
Kraus, K., & Pfeifer, N. (2001). Advanced DTM generation from LIDAR data.
International Archives of Photogrammetry and Remote Sensing, XXXIV, 23–30.
23
Lan, P. H. (2009). Building a unified height system with various initial points for the
ASEAN countries. In 7th FIG regional conference. Spatial data serving people.
Hanoi, Vietnam: FIG, p. 9.
Land Registration Act 2002 (2002). An act to make provision about land
registration; and for connected purposes. <http://www.legislation.gov.uk>
26.02.2.
Lidberg, M., Johansson, J. M., Scherneck, H.-G., & Davis, J. L. (2007). An improved and
extended GPS-derived 3D velocity field of the glacial isostatic adjustment (GIA)
in Fennoscandia. Journal of Geodesy, 81, 213–230.
Meier, E., Geiger, A., Ingensand, H., Licht, H., Limpach, P., Steiger, A., et al. (2011).
Hydrostatic leveling systems: Measuring at the system limits. FIG congress. Sydney,
Australia: Springer (p. 15).
Navratil, G. (2008). Legal and technical aspects of decisions on property boundaries
– The case of Austria. Nordic Journal of Surveying and Real Estate Research, 5,
7–23.
Neunzert, G. M. (2011). Subdividing the land – Metes and bounds and rectangular
survey systems. Boca Raton, FL: CRC Press.
Stoter, J., & van Oosterom, P. (2006). 3D Cadastre in an international context. Boca
Raton: Taylor & Francis.
Twaroch, C. (2005). Richter kennen keine Toleranz (engl. Judges do not use
Imprecision). In G. Chesi & T. Weinold (Eds.), Int Geodätische Woche
(pp. 154–159). Obergurgl: Wichmann.
van Oosterom, P. J. M., Stoter, J. E., & Lemmen, J. E. (2005). Modelling of 3D cadastral
systems. In K. Kyu-Tae (Ed.), 28th Cadastral seminar (pp. 594–606). Busan: Korea
Cadastral Survey Corp..
VwGH (1983). Erkenntnis des Verwaltungsgerichtshofes (engl. Finding of the
Austrian Highest Administrative Court). 83/06/0088.
Waalenwijn, I. A. (1964). Hydrostatic levelling in the Netherlands. Survey Review,
17(212–221), 267–276.
Wolf, A. (1940). Tidal force observation. Geophysics, 5, 317–320.
Wolf, P. R., & Brinker, R. C. (1989). Elementary Surveying (8th ed.). New York: Harper
& Row.