GPS ALIGNMENT SURVEYS AND

MERIDIAN CONVERGENCE
By Tomás Soler,1 Member, ASCE, and Rudolf J. Fury2

ABSTRACT: Since the advent of the Global Positioning System (GPS), geodetic
azimuths can be accurately computed by simple implementation of well-known 3D
concepts. However, when GPS alignment surveys involving azimuths are designed
in advance, and later observed and reduced (e.g., during kinematic GPS work),
corrections due to the convergence of the meridians should be kept in mind and
not ignored. In this study a practical algorithm was used to compare accurately
determined ‘‘meridian convergence’’ against the classical formalism available in
standard textbooks. The typical approximate formulation available in the open lit-
erature was found adequate for GPS engineering surveys such as airport runway
profiles, alignment of power lines or conveyor belts, stake positioning in highway
construction, etc. A practical GPS survey was used to corroborate the results. Fi-
nally, a new 3D alternative to computing meridian convergence, which is equivalent
to the rigorous formalism, is presented.

INTRODUCTION
Meridian convergence can be defined in three ways (Geodetic 1986, p.
159):

• On a curved surface (globular body) it is the mutual approach of the

meridians in passing from the equator to the poles. At the equator all
meridians are parallel. Extending poleward from the equator, they draw
together until they meet at the poles, intersecting in angles equal to their
difference of longitude.
• On a curved surface (globular body) and for a given pair of meridians
and a given geodesic, it is the difference between the two angles formed
by the intersection of the geodesic with the two meridians. Thus the
azimuth at one end of the geodesic differs from the azimuth at the other
end by 180⬚ plus or minus the amount of the convergence of the merid-
ians at the endpoints.
• On a map, at a given point, it is the angle measured clockwise from the
tangent to the projection of the meridian to the northing coordinate line
(grid north)—also called map angle or mapping angle.

The first definition is the most general and the one applicable when Global
Positioning System (GPS) methodology is employed. Thus, the geodetic az-
imuths change between pairs of points A and B, B and C, . . . , F and G, etc.,
even when all of these points are spanning the same straight line AG. From
the above definitions we know that convergence only makes sense between
meridians on the surface of the ellipsoid and that the convergence is zero at

1
Chf., Global Positioning Sys. Branch, Nat. Geodetic Survey, NOS, Nat. Oceanic
and Atmospheric Admin., 1315 East-West Hwy., Silver Spring, MD 20910.
2
Geodesist, Nat. Geodetic Survey, NOS, Nat. Oceanic and Atmospheric Admin.,
1315 East-West Hwy., Silver Spring, MD.
Note. Discussion open until January 1, 2001. To extend the closing date one month,
a written request must be filed with the ASCE Manager of Journals. The manuscript
for this paper was submitted for review and possible publication on June 28, 1999.
This paper is part of the Journal of Surveying Engineering, Vol. 126, No. 3, August,
2000. 䉷ASCE, ISSN 0733-9453/00/0003-0069–0082/$8.00 ⫹ $.50 per page. Paper
No. 21304.

JOURNAL OF SURVEYING ENGINEERING / AUGUST 2000 / 69

the equator and equal to the difference in longitude of the two points if they
are close to the poles. The third definition refers to the convergence between
the representation of meridians as projected using different mapping equa-
tions. This convergence will depend on the particular cartographic projection
at hand and will not be discussed further in this paper. However, it can be
said that meridian convergence for regular cylindrical projections is zero, and
for regular conic and polar azimuthal projections it is a simple function of
latitude. Other projections have a more complicated set of equations affected
by their intrinsic distortions and are readily available in the geodetic litera-
ture. As we will see, the meridian convergence may become several minutes
of arc in most surveying applications and thus must not be ignored. It is
easily detectable with modern GPS technology and can be significant even
when the surveyed points are located at short distances from each other.
Despite this fact, the meridian convergence can be approximated to sufficient
accuracy by simple equations.

APPROXIMATIONS TO MERIDIAN CONVERGENCE

Approximation No. 1
An illustration of the first definition is depicted graphically in Fig. 1(a).
The mathematical reasoning followed is the usual treatment of the subject in
most classical textbooks [e.g., Zakatov (1962, p. 100)]. The convergence
between any two points A and C located at the same geodetic latitude ␾A is
visualized by drawing tangents to the ellipsoid along the meridians of A and
C until they intersect the semiminor axis of the ellipsoid at E. By definition,
the angle dt between the tangents is the so-called convergence of the merid-
ians going through points A and C. The convergence between two arbitrary
nearby points A and B, not necessarily on the same parallel [Fig. 1(a)], could

FIG. 1. Meridian Convergence on Surface of Ellipsoid (Diagram Not Drawn to

Scale)

70 / JOURNAL OF SURVEYING ENGINEERING / AUGUST 2000

be determined only by assuming a differential local area around point A.
Because of this restriction some planar approximations can be enforced. This
simplified theory is more than sufficient to cover all usual surveying and
surveying-engineering applications. Consequently, the formalism delineated
herein will confine the applicability of the equations presented to cases in
which the points are located in their local vicinity, not exceeding a distance
from point A of more than 10-km radius. If longer distances are anticipated,
the simplifications introduced will no longer be valid, and the second defi-
nition should be considered.
A basic but important geometric relationship is
arc (linear units) = central angle (radians) ⫻ circular radius (linear units)
Applying the above equality to triangle EAC in Fig. 1(a) and also using Fig.
1(b), we can write
arc AC = dtNA cot ␾A (1)
Similarly, from triangle FAC we have
arc AC = d␭NA cos ␾A (2)
Because the arcs in (1) and (2) are approximately equal, it immediately fol-
lows that
dt = d␭ sin ␾A (3)
For the differential planar case, an element of geodesic ds is equal to an
element of normal section and is a straight line [Fig. 1(c)]. From triangle
ABC, we obtain
arc AC = ds sin ␣AB (4)
Equating (4) and (2), an expression for d␭ is found
d␭ = ds sin ␣AB /(NA cos ␾A) (5)
Finally, substituting (5) into (3) provides the formula to compute the con-
vergence of meridians dt as a function of the geodetic latitude ␾A of point
A, the differential distance ds between the two points, and the azimuth ␣AB
from point A to point B, namely
dt = ds sin ␣AB tan ␾A /NA (6)
The symbol NA refers to the principal radius of curvature of the prime vertical
of point A, computed as follows: NA = a/W with W = (1 ⫺ e2 sin2␾A)1/2,
where the square of the ellipsoid eccentricity is given in closed form by e2
= 2f ⫺ f 2. The ellipsoid of revolution adopted in this investigation is the
GRS80 ellipsoid (Moritz 1992); its two geometric parameters of interest are
size, defined by the semimajor axis a = 6,378,137 m, and shape, specified
by the flattening f or, equivalently, its inverse, f ⫺1 = 298.257222101.
Eq. (6) will be referred to in the text as Approximation No. 1 for deter-
mining the meridian convergence.

Approximation No. 2
An alternative and better approximation to meridian convergence was in-
troduced by the U.S. Coast and Geodetic Survey (Formulas 1933, pp. 8 and
97). The final equation is given as
JOURNAL OF SURVEYING ENGINEERING / AUGUST 2000 / 71
 dt = d␭ sin 1/2(␾m)sec 1/2(␾B ⫺ ␾A) ⫹ (d␭)3 sin ␾m cos2␾m /12 (7)
where ␾m = 1/2(␾A ⫹ ␾B) = mean latitude of points A and B. The value of
d␭ required in the above equation was previously given in (5).

RIGOROUS MERIDIAN CONVERGENCE

Geodetic Azimuth ␣AB
Now, attention will be paid to the unambiguous definition of azimuth ␣AB.
To clarify the symbolism, two similar, but distinct, definitions are reviewed:

• Geodesic azimuth ␣⬘AB —the azimuth of the geodesic between two points
A and B on the surface of the reference ellipsoid measured at A.
• Geodetic azimuth ␣AB —the azimuth between two spatial points A and
B (not necessarily on the ellipsoid) measured clockwise from the geo-
detic north on the geodetic horizon plane of A, this plane being at a
right angle to the ellipsoidal normal. This parameter is referred to by
some authors as 3D azimuth (Burkholder 1997).

When points A and B are near each other (ⱕ10 km), ␣AB ⬃ ␣⬘AB. Conse-
quently, a unique notation ␣AB and the definition of geodetic azimuth will be
assumed throughout the remainder of this work. It is well known that geodetic
azimuths can be related to astronomic azimuths (a physical quantity) through
the components of the deflection of the vertical and the so-called Laplace’s
equation or condition (Heiskanen and Moritz 1967, pp. 186 and 190). A
combination of GPS and geodetic leveling observations also permits the ac-
curate determination of vertical deflections (Soler et al. 1989).
Geodetic azimuths between any two arbitrary points are the only ones that
can be determined by 3D GPS techniques and methods and thus are the only
ones of interest in current geodetic and surveying GPS practice. However,
the definition of geodetic azimuth makes more sense if the two points are
intervisible, this being independent of the distance between such a pair of
points. The same interpretation does not apply to geodesic azimuths that are
strictly referred to points lying on the surface of the ellipsoid. Hence, a
geodetic azimuth between points on opposite sides of the reference ellipsoid,
although computable, does not usually have practical interest. However, de-
termining instantaneous geodetic azimuths of visible celestial objects or ar-
tificial satellites is perfectly sound and sometimes necessary. A practical ex-
ample related to communication geostationary satellites was presented in
Soler and Eisemann (1994).
Recalling Fig. 1(a), notice that we have drawn the local geodetic frame
[i.e., (e, n, u)] at point A. Local geodetic frames are commonly referred to
in the geodetic literature as east-north-up frames (Soler and Hothem 1988).
The particular notation used here is consistent with right-handed coordinate
frames, the only ones considered throughout this paper.
The typical data output in a GPS survey contains, among other quantities,
the components of vectors (⌬xAB, ⌬yAB, ⌬zAB) between points (e.g., A and
B). However, these components are given with respect to a local (x, y, z)
conventional terrestrial frame. The (x, y, z) frame is parallel to the geocentric
conventional terrestrial reference frame (x, y, z) (see Appendix II for defi-
nitions). The conversion between local conventional terrestrial frames (x, y,
z) and local geodetic frames (e, n, u) is accomplished by applying a trans-
formation rotation matrix [R], which enforces parallelism between those
frames.
72 / JOURNAL OF SURVEYING ENGINEERING / AUGUST 2000
 The geodetic azimuth ␣AB and ‘‘geodetic altitude’’ ␯AB (i.e., geodetic ver-
tical angle) between two points A and B can be computed using the following
two equations:
tan ␣AB = eAB /nAB (8)
tan ␯AB = uAB /(e 2
AB ⫹n )2
AB
1/2
(9)
This implies that the components of the vector AB (eAB, nAB, uAB) must be
known along the local geodetic coordinate frame (e, n, u). These can be
calculated by transforming the components (⌬xAB, ⌬yAB, ⌬zAB) along the (x,
y, z) frame at A as given by GPS, into the (e, n, u) frame. This transformation
is performed by using the rotation matrix [R]A defined as follows [e.g., Soler
(1976) and Soler and Hothem (1988)]:
[R]A = R1(1/2␲ ⫺ ␾A)R3(␭A ⫹ 1/2␲) = R3(1/2␲)R2(1/2␲ ⫺ ␾A)R3(␭A)

= 冋 ⫺sin ␭
⫺sin ␾ cos ␭
cos ␾ cos ␭
cos ␭
⫺sin ␾ sin ␭
cos ␾ sin ␭
0
cos ␾
sin ␾
册 A (10)
and the matrix equation

再冎 再 冎
eAB
nAB
uAB
= [R]A
⌬xAB
⌬yAB
⌬zAB
(11)

If the components (⌬xAB, ⌬yAB, ⌬zAB) are not immediately known, they can
be determined from the geocentric Cartesian coordinates of points A and B

再 冎 再冎 再冎
⌬xAB
⌬yAB
⌬zAB
=
xB
yB
zB
⫺
xA
yA
zA
(12)

The rectangular coordinates, x, y, z of any arbitrary point can also be com-

puted from their known curvilinear geodetic coordinates (␭, ␾, h) employing
well-known expressions

再冎 再
x
y
z
=
(N ⫹ h)cos ␾ cos ␭
(N ⫹ h)cos ␾ sin ␭
[N(1 ⫺ e2) ⫹ h]sin ␾
冎 (13)

where N, as mentioned above, is the principal radius of curvature in the prime

vertical. A noniterative inverse transformation of (13) to compute geodetic
coordinates from Cartesian coordinates based on the work of Bowring (1985)
was published in Soler and Hothem (1989).

Numerical Investigation
To ascertain the accuracy of (6) and (7) when used as an alternative for
the approximation of meridian convergence, a numerical simulation was per-
formed. It is based on a rigorous 3D approach that emulates the logic applied
in everyday GPS work.
A horizontal line joining two arbitrary stations A and B separated by a
distance of 10 km was selected to restrict the outcome of the analysis to
practical engineering surveys. Because the expected results are azimuth de-
pendent, the values of ␣AB were increased by increments of 15⬚ from the
meridian of point A. Similarly, to account for the dependence of convergence
JOURNAL OF SURVEYING ENGINEERING / AUGUST 2000 / 73
FIG. 2. Meridian Convergence Along Line AB of Azimuth ␣AB (Diagram Not
Drawn to Scale)

FIG. 3. Algorithm to Rigorously Determine Meridian Convergence

on distance, intermediate points between A and B were introduced having a

constant spacing of 1 km between them. Recall that from (6) the convergence
dt is also a function of the latitude of the initial point A where the original
azimuth ␣AB is measured; thus, to identify exactly the two points between
which the convergence is measured, a more explicit notation dtAB, indicating
the convergence at B between the meridians of A and B, is introduced (Fig.
2). Similarly, we write dtAP⬘, dtAQ⬘, etc. The test performed in this numerical
investigation was based on an a priori selection of geodetic coordinates for
point A, specifically, ␭A = 100⬚W, ␾A = 40⬚N; hA = 0, which places A on the
reference ellipsoid near the center of the conterminous United States.
The flowchart of the rigorous algorithm used to compute the convergence
of meridians between two points A and B⬘ on the ellipsoid and its graphical
74 / JOURNAL OF SURVEYING ENGINEERING / AUGUST 2000
 FIG. 4. Schematic Diagram to Rigorously Compute Meridian Convergence

interpretation is recounted step-by-step in Figs. 3 and 4. Notice that P⬘ is the

projection of P on the surface of the ellipsoid (hP⬘ = 0). Thus all the com-
putations are performed along tangents to the surface of the ellipsoid at points
A, P⬘, Q⬘, etc. It should be remembered that the normals to the ellipsoid at
P⬘, Q⬘, R⬘, etc., are not generally contained on the same plane. Exceptions
include points along the same meridian or at the equator where the conver-
gence is zero. Thus, the azimuths (e.g., ␣AB and ␣P⬘Q ) are not measured on
the same local geodetic plane. However, because the distances AP, P⬘Q, etc.
(Fig. 4) are only 1 km, the error introduced is negligible. This approach is
more rigorous than computing the meridian convergence using the equality
dtAB = ␣BA ⫺ (␣AB ⫹ ␲) (14)
because the distance between A and B is 10 km and the local geodetic planes
at A and B are not parallel. Our simulation shows that although the ellipsoidal
height of point B is 7.8 m (distance BB⬘ in Fig. 4) the difference in conver-
gence computed using the rigorous algorithm and (14) is only 0⬙.02. Thus
for all practical purposes, once the coordinates of P⬘ are known (Fig. 2), the
meridian convergency dtAP⬘ can be determined using (14).
The rigorous meridian convergence as a function of distance and azimuth
between two arbitrary points with the restrictions mentioned above is shown
in Fig. 5. The convergence of the meridians is significant even at short dis-
tances (e.g., 100 arc seconds at 5 km and azimuth of 45⬚) and, if neglected,
can easily accumulate with distance. The convergence between meridians
increases with the value of azimuth, reaching a maximum at azimuths of 90⬚.
This could be explained by simple geometric principles. Restricting the study
to the first quadrant, and assuming a constant distance from point A, the
separation between the meridians of the two points, which is directly pro-
portional to the convergence, increases with azimuth. Fig. 5 is strictly valid
only for points of ␾ = 40⬚. Considering that the parallels increase in curvature
per unit length approaching the north pole, the meridians get closer and closer
for a constant interval of longitude, Fig. 6 shows the value of the meridian
convergence for a constant azimuth ␣ = 45⬚ at various latitudes.
It should be emphasized that the magnitude of the convergence, although
JOURNAL OF SURVEYING ENGINEERING / AUGUST 2000 / 75
FIG. 5. Rigorous Meridian Convergence as Function of Distance and Azimuth

FIG. 6. Meridian Convergence at Different Latitudes ␾ and Constant Azimuth

␣ = 45ⴗ (Constant)

76 / JOURNAL OF SURVEYING ENGINEERING / AUGUST 2000

FIG. 7. Difference of Meridian Convergence between Symmetric Northern
and Southern Points with Respect to 90ⴗ Azimuth

FIG. 8. Graphic Display of Meridian Convergence at Northern and Southern

Points (Diagram Not Drawn to Scale)

JOURNAL OF SURVEYING ENGINEERING / AUGUST 2000 / 77

 FIG. 9. Differences in Meridian Convergence (Rigorous-Approximate)

symmetric with respect to longitude due to the rotational symmetry of the

ellipsoid, is not likewise symmetric with respect to latitude on account of the
changing north-south curvature of the ellipsoid. Fig. 7 depicts the differences
in meridian convergence for points symmetrically located with respect to the
east-west line of point A. The maximum difference in convergence between
northern and southern points happens at azimuths of 45⬚ and 135⬚ and
amounts to about only 0.15 arc seconds at distances of 5 km. A conclusion
not obvious beforehand can be drawn. Because of the symmetry implicit in
the geometry of the problem, the differences between the northern and south-
ern values of the convergence are almost identical for the pairs 15⬚ versus
165⬚ and 75⬚ versus 105⬚, and 30⬚ versus 150⬚ and 60⬚ versus 120⬚. Fig. 8
is a graphical interpretation of this interesting finding, which displays two
cases of the example studied.
When the approximate equations [(6) and (7)] were compared to the rig-
orous approach, the differences encountered up to distances of 10 km are not
significant. This is depicted in Fig. 9, which shows that the discrepancy be-
tween the rigorous algorithm and the two approximate equations is maximum
at 45⬚ azimuths. Refined (7) gives slightly better results. The graph shows
that, at 45⬚ azimuths and distances of 10 km, the maximum difference be-
tween the rigorous algorithm and (6) reaches only 0.30 arc seconds. The
actual convergence at this distance and azimuth is 180 arc seconds (Fig. 5).
This implies that, in the worst possible scenario, we are making only a 0.2%
error when we approximate the convergence of meridians by (6). For dis-
tances <6 km, the difference between the approximate equation and the rig-
78 / JOURNAL OF SURVEYING ENGINEERING / AUGUST 2000
 FIG. 10. General 3D Meridian Convergence (Diagram Not Drawn to Scale)

orous numerical approach does not exceed 0.1 arc seconds. This difference
is smaller than the precision of most theodolites used in surveying engineer-
ing field work. Another characteristic of Fig. 9 shows that the differences
between the rigorous algorithm and (6) and (7) are symmetric with respect
to 45⬚ azimuths and zero at 90⬚ azimuths. These are consequences of the
assumptions implicit in the simplifications of the approximate equations. For
example, for short distances and azimuths of 90⬚ the three approaches give
the same answer for the value of the meridian convergence. In Fig. 1(a), arc
AC will approximate a straight line, and the three approaches to compute
meridian convergence are equivalent.
It is important to recognize that in surveying practice it is incorrect to
assume that the azimuth from P⬘ to B (Fig. 2) with P⬘ on the line AB has
the same azimuth ␣AB as the line AB itself. The convergence of meridians is
significant even for distances typically involved in precise surveying align-
ments and should never be ignored.
Once the computations on the ellipsoid were implemented according to the
logic described, a more general 3D approach was investigated. Fig. 10 shows
that by the definition of geodetic azimuth between two points A and B, the
azimuth ␣AB at A is equal to the azimuth ␣A⬘B from A⬘ (the projection of A
on the surface of the ellipsoid) to B measured on the geodetic horizon plane
of A⬘. This results from the definition of local coordinate systems (e, n, u),
which are parallel at A and A⬘. Thus, the intersection of the plane ABDA⬘
with the local geodetic horizon planes of A and A⬘ produces the same geo-
detic azimuths, in other words, ␣AB = ␣A⬘B. What then is the convergence
dtAB ? By definition it is the spatial angle between the n-axis at B and an axis
parallel to the n-axis of A at point B (Fig. 10).
Based on the above definition, the convergence dtAB can be approximately
written as follows:
JOURNAL OF SURVEYING ENGINEERING / AUGUST 2000 / 79
 FIG. 11. Algorithm to Compute 3D Meridian Convergence

FIG. 12. Circular Cumulative Station Offsets due to Neglecting Meridian Con-
vergence Errors

dtAB = ␣BA⬙ ⫺ ␣AB (15)

where the point A⬙ is the tip of a vector BA⬙ that has the same orientation and
magnitude as the vector AB. The approximation involved is the assumption
that the axis through B parallel to the n-axes at A is in the local geodetic
horizon of B. The algorithm used to compute the convergency anywhere in
80 / JOURNAL OF SURVEYING ENGINEERING / AUGUST 2000
space is presented in Fig. 11. We restricted the computations to points that are
<10 km apart, but the same logic can be expanded to longer distances if
necessary. Restricting ourselves to distances of a maximum of 10 km, the
difference between the rigorous ellipsoidal and 3D algorithms only amounted
to 0⬙.007 when ␯AB = 5⬚. Because of its generality and geometric simplicity,
the 3D approach to compute meridian convergence is recommended in practice.

FIELD EXAMPLE
To appreciate the practical implications of neglecting the meridian con-
vergence, a real example will be discussed. The test area selected is located
at the Municipal Airport of Newark, Ohio. The two ends of a typical runway,
points A and B, are accurately surveyed using static GPS procedures and ties
to nearby points of the National Continuously Operating Reference Stations
(CORS) network. Then, a kinematic GPS survey is performed starting at
runway endpoint A, reaching B, and returning to A to provide redundancy
and to check for blunders. The principal objective of the survey is to have
an accurate geodetic height profile between points A and B along the line
defining the center of the runway. An interpolation routine was developed to
determine the average height of the two measurements along the line AB at
intervals of 50 m. The locations of the equidistant points along route AB
were computed using the coordinates of the initial point A and the azimuth
from A to B, which could be accurately computed from the coordinates of A
and B. Not correcting this azimuth for the effects of meridian convergence
(the rigorous algorithm was used) at each 50-m interval results in the accu-
mulation of circular positional errors depicted in Fig. 12. When performing
surveying alignments of distances even as short as 1,400 m—the actual
length of this particular runway—and for an azimuth close to 90⬚ a closing
error of about 13 cm (almost 1/2 ft) would be accumulated by not correcting
for meridian convergence at each 50-m section.

CONCLUDING REMARKS
The correction for meridian convergence is critical when computing geo-
detic coordinates of points that should be accurately aligned between two
known stations A and B by traversing intermediate points P⬘, Q⬘, etc. As Fig.
5 shows, meridian convergence depends on the location of the initial point
A, the geodetic azimuth ␣AB , and the distance between A and B. Thus, (6)
or (7) are used to approximate the meridian convergence when computing
coordinates for planning alignment projects. If preferred, a new more rigorous
3D algorithm could be implemented to determine the value of the meridian
convergence between any two points in space, although its practicality is
restricted to topographic mapping applications. It is important to realize that
meridian convergence errors can be detected using modern GPS techniques.
Neglecting this correction may introduce inadmissible errors when calculating
the coordinates of points along profiles that must be positioned in the field
using traversing GPS methods. Hence, meridian convergence becomes sig-
nificant for distances exceeding 1 km and should be taken into account when
rigorous GPS surveying alignments are established.

ACKNOWLEDGMENTS
The writers appreciate the contribution of their National Geodetic Survey colleague Jeff
Olsen, whose inquiry prompted this investigation. The writers are also indebted to an anonymous
reviewer for helpful comments.

JOURNAL OF SURVEYING ENGINEERING / AUGUST 2000 / 81

APPENDIX I. REFERENCES
Bowring, B. R. (1985). ‘‘The accuracy of geodetic latitude and height equations.’’
Survey Rev., 28(October), 202–206.
Burkholder, E. F. (1997). ‘‘Three-dimensional azimuth of a GPS vector.’’ J. Surv.
Engrg., ASCE, 123(4), 139–146.
Formulas and tables for the computation of geodetic positions. (1933). U.S. Coast
and Geodetic Survey Spec. Publ. No. 8. U.S. Government Printing Office, Wash-
ington, D.C.
Geodetic glossary. (1986). National Geodetic Survey, Silver Spring, Md.
Heiskanen, W. A., and Moritz, H. (1967). Physical geodesy, W. H. Freeman and Co.,
San Francisco; Reprinted (1981), Inst. of Phys. Geodesy, Technical University,
Graz, Austria.
Moritz, H. (1992). ‘‘Geodetic reference system 1980.’’ Bull. Géodésique, 66(2), 187–
192.
Soler, T. (1976). ‘‘On differential transformations between Cartesian and curvilinear
(geodetic) coordinates.’’ Rep. No. 236, Dept. of Geod. Sci., Ohio State University,
Columbus, Ohio.
Soler, T., Carlson, A. E., Jr., and Evans, A. G. (1989). ‘‘Determination of vertical
deflections using the global positioning system and geodetic leveling.’’ Geophys.
Res. Letter, 16(7), 695–698.
Soler, T., and Eisemann, D. W. (1994). ‘‘Determination of look angles to geostationary
communication satellites.’’ J. Surv. Engrg., ASCE, 120(3), 115–127.
Soler, T., and Hothem, L. D. (1988). ‘‘Coordinate systems used in geodesy: Basic
definitions and concepts.’’ J. Surv. Engrg., ASCE, 114(2), 84–97.
Soler, T., and Hothem, L. D. (1989). ‘‘Important parameters used in geodetic trans-
formations.’’ J. Surv. Engrg., ASCE, 115(4), 414–417.
Zakatov, P. S. (1962). A course in higher geodesy, National Technical Information
Service, U.S. Department of Commerce, Washington, D.C. (translated from Rus-
sian).

APPENDIX II. NOTATION

The following symbols are used in this paper:

(e, n, u) = local (right-handed) geodetic coordinate system at any point

(␭, ␸, h); e-axis points to (geodetic) east, n to (geodetic)
north, and u to (geodetic) zenith;
h = geodetic height (i.e., height above ellipsoid);
N = principal radius of curvature in prime vertical plane;
(x, y, z) = local (terrestrial) frame; origin is at point of observation and
x-, y-, and z-axes are parallel to x-, y-, and z-axes, respec-
tively;
(x, y, z) = conventional terrestrial reference frame; Earth’s fixed geo-
centric coordinate system; z points toward conventional ter-
restrial pole, x passes through point of zero longitude as de-
fined by International Earth Rotation Service; y forms
right-handed coordinate system with x and y [e.g., ITRF97,
epoch 1997.0; WGS84 (G873), epoch 1997.0];
␣ = geodetic azimuth;
␣⬘ = geodesic azimuth; and
(␭A , ␾A , hA ) = curvilinear geodetic coordinates (longitude, latitude, and el-
lipsoid height) of an arbitrary point A.

82 / JOURNAL OF SURVEYING ENGINEERING / AUGUST 2000