Fachbeitrag Lars E Sjberg, New Solutions to the Direct and Indirect Geodetic Problems on the Ellipsoid

131. Jg. xxx/2006 zfv

Summary
Taking advantage of numerical integration solves the direct
and indirect geodetic problems on the ellipsoid. In general
the solutions are composed of a strict solution for the sphere
plus a correction to the ellipsoid determined by numerical in-
tegration. The solutions are demonstrated by three numeri-
cal examples.
Zusammenfassung
Die klassische Berechnung der geodtischen Hauptaufgaben
auf dem Rotationsellipsoid beruht auf der Entwicklung der In-
tegranden der elliptischen Integrale in Taylor-Reihen. In die-
sem Beitrag wird die exakte Lsung fr die Kugel mit Kor-
rekturen fr das Ellipsoid, die durch numerische Integration
bestimmt werden, kombiniert. Als eine weitere Anwendung
kann die Flche eines geodtischen Polygons mit dieser Me-
thode berechnet werden. Die Lsungen werden anhand von
drei numerischen Beispielen veranschaulicht.
Introduction
The direct and indirect geodetic problems on the ellipsoid
have attracted the attention of numerous geodesists in
the past. Although ellipsoidal solutions are of less impor-
tance in the space age, problems related with geodesics
may still be relevant. Examples can be found e. g. in the
application of the Law of the Sea, military surveys and in
the optimisation of aircrafts and ships routes.
The classical solutions of these problems were fre-
quently in the forms of power series of the ellipsoidal
eccentricity. Some of the basic formulas used below are
provided e. g. by Heck (1987). Today it is natural to take
advantage of numerical integration by computers. Klotz
(1991) and Schmidt (1999) presented some pioneer work
along this line, and our solutions to the direct and indirect
problems will also rely on numerical integrations. Sim-
ilarly, the area under the geodesic can be conveniently
computed by numerical integration (Sect. 4).
2 Thedirectproblemontheellipsoid
Let us assume that the ellipsoidal parameters a and b, the
semi-major and -minor axes, are given. From these we
compute the first eccentricity by
2 2
e a b / a = . More
details on the basic formulas used are presented in Heck
(1987), Klotz (1991) and Schmidt (1999).
Problem: Let us assume that the coordinates of latitude
( )
1
and longitude ( )
1
of point P
1
are given. Also given
are the azimuth
1
at P
1
and the geodesic arc length s
to another point P
2
. Determine the coordinates of P
2
(see
Fig. 1).
Solution:
a) Let us first convert some trigonometric functions of the
geodetic latitude to functions of the reduced latitude :
tan k tan = ,
2 2
sin k sin / 1 e sin = and
2 2
cos cos / 1 e sin = (1)
where
2
k 1 e = .
b) Compute Clairauts constant for the geodesic by the
formula
1 1
c cos sin = . (2)
If
1 2
and starting from the following equation for
the given arc length s (Heck 1987 and Schmidt 1999):
2
1
2 2
2 2
1 e cos
s a cos d
cos c

, (3a)
we arrive at the formula
2
1
x x
2
x x
x
s a arcsin ds
1 c
=
=
l
l
=
l

l

, (3b)
where
( )
( )
2
1
x 2
2
2 2 2 2 2
x
1 x dx
ds ae
1 c x 1 1 e e x

. (3c)
New Solutions to the Direct and Indirect
Geodetic Problems on the Ellipsoid
Lars E Sjberg
1

s
2 2 2
P ( , ) ?
1 1 1
P ( , )
North
Fig.:Thedirectproblemontheeelipsoid
Fachbeitrag L. E. Sjberg, New Solutions to the Direct and Indirect Geodetic Problems on the Ellipsoid
2 zfv xxx/2006 131. Jg.
Hence,
2
can be determined by
( )
( )
2 2
2 1
sin 1 c sin s ds / a arcsin x / 1 c
l
=
l
l
, (4)
where ds is preferably determined by numerical integra-
tion. As the upper integration limit depends on
2
, Eq. (4)
should be iterated. As a starting value ds can be set to 0.
[One problem of Eq. (3c) is the singularity of the integral
for
2
x 1 c = .]
If
1 2
= , of course, the above procedure does not hold,
but
1 2
= can be checked by the formula
1
1
2
x
s a 2arcsin ds
1 c

\

=

( )

, (5a)
where
( )
2
1
1 c
2
2
1
2 2 2 2 2
x
(1 x )dx
ds 2ae
1 c x 1 1 e e x

. (5c)
Remark: If
1 2
the iteration of Eqs. (3b) and (3c) may
be very slow or fail, which calls for a special investiga-
tion out of the scope of this article.
c) If
1 2
, the longitude
2
is given by the formula
(e. g. Heck 1987 or Schmidt 1999)
2
1
2
1
2 2
2 1 2 2
1
2 2
i 2
i
1
0
i 1
1 e cos d
c
cos c cos
d
c d
cos cos c
t
arcsin d
t

=
=

l
l
=
l
l

(6a)
where
( )
2 2
1 1
x sin
2
2 2 2 2 2
x sin
dx
d e c
1 c x 1 1 e e x

=
=
=

, (6b)
i i
t tan = and
2
0
t 1 c / c = .
The integral of Eq. (6b) can be determined by numerical
integration between the limits x
1
and x
2
.
If
1 2
0 = , the longitude of the new point is given by
( )
2
1
1 c
2 1
2 2 2
x
1
1
0
dx
2c d '
1 x 1 c x
t
2 arcsin d '
2 t

=

l
l
=
l
l

(7a)
where
( )( )
2
1
1 c
2
2 2 2 2 2
x
dx
d ' 2e c
1 c x 1 1 e e x

. (7b)
As an alternative to the above equations we may take ad-
vantage of the similarities of Eqs. (3a) and (6a) to derive
1
c
s I d
a
= , (8a)
where
( )
( )
( )
2
1
2
1
i 2 x
2
i i
1 2
2 2 2 2
0
x i 1
1 c
2
1 1
1 2
2 2 2 2
0
x
x dx t x
c arcsin carcsin , x x
t
1 x 1 c x 1 c
I
x dx x t
2c 1 c 2 carcsin arcsin , x x .
t
1 x 1 c x 1 c

=
=

'
1 \
1
1 =
1

1
( )
1
11
= !
1
\ 1
1
= = 1

1

( ) 1
11+

(8b)
and
( )
( )
2
1
2
1
x
2
2
1 2
2 2 2 2 2
x
1
1 c
2
2
1 2
2 2 2 2 2
x
x dx
ce if x x
1 c x 1 1 e e x
d
x dx
2ce if x x .
1 c x 1 1 e e x

'
1
1
1
1
1

1
1
1
1
=
!
1
1
1
=
1
1
1
1
1+1

(8c)
In the special case that both given points are located on
the equator Eq. (8a) reduces to
( )
( )
2
1 c
2
2
2 2 2 2 2
0
cs x dx
1 c 2ce
a
1 c x 1 1 e e x


=

,
(9)
or, approximately,
2
1 c
2
2
2 2
0
2
2
cs x dx
(1 c) ce
a
1 c x
cs c(1 c )
(1 c) e
a 4

. (10)
This completes the solution of the direct problem on the
ellipsoid.
3 Theindirectproblem
In the indirect problem the coordinates of two points (say
P
1
and P
2
) are given.
Problem: Determine the arc length along the geodesic be-
tween the two points and the azimuths of the geodesic at
the two points.
Fachbeitrag Lars E Sjberg, New Solutions to the Direct and Indirect Geodetic Problems on the Ellipsoid
3 131. Jg. xxx/2006 zfv
Solution:
a) First convert the geodetic latitudes
i
to reduced lati-
tudes
i
.
b) Then determine Clairauts constant for the geodesic
running through the two given points (Sjberg 2005b).
Three cases must be distinguished.
i) In the general case with
1 2
, compute

2 2 2 2
2 1 2 1 2
0 2
t t 2t t cos DL
t
sin DL

= , (11a)
where
i i
t tan = ,

2 1 c
DL d = (11b)
and

( )
2
1
x
2
c
2 2 2 2 2
x
dx
d e c
(1 c x ) 1 1 e e x
=

. (11c)
Eq. (11c), with
i i
x sin = , is conveniently integrated
numerically. Clairauts constant is then given by

2
0
c 1/ 1 t = . (12)
ii) If
1 2
0 = , Clairauts constant is given by the
equation

( )
( )
2 2
1
cos D
c
t cos D

, (13a)
where (with
1 1
x sin = )

( )( )
2
1
1 c
2 2 1
2 2 2 2 2
x
dx
D e c
2
1 c x 1 1 e e x

.


(13b)
iii) If
1 2
0 = = , Clairauts constant is given by

( )
2 2
11
cos DL
c
t cos (DL)
=

, (14a)
where
11 11
t tand = ,
11 max
d < being an arbitrary
but small latitude, and

( )( )
2
1 c
2 2 1 11
2 2 2 2 2 2
0
dx cd
DL e c
2
1 c 1 c x 1 e e x

=


.


(14b)
In all three cases the procedures require iteration to reach
the final determination of c.
c) The arc length follows from Eqs. (3a) and (3b). If
1 2
= , we may take advantage of Eqs. (8a)-(8c) to derive
( )
1
a
s I d
c
= . (15)
In particular, if
1 2
0 = = , the arc length can be ap-
proximated by
( )( )
2
2
a e
s (1 c) 2 c 1 c
c 4

l
l

l
l
. (16)
d) The azimuths at points P
1
and P
2
are given by Clairauts
equation, Eq. (2):
i i
sin c/ cos = . (17)
This completes the solution to the indirect geodetic prob-
lem on the ellipsoid.
4 Areacomputationunderthegeodesic
Following Sjberg (2005a) the area (T) on the ellipsoid
limited by the equator, a geodesic with the Clairaut con-
stant c and the two meridians with longitudes
1
and
2 1
can be written
[ [
2 2
1 1
2 2 2
2 1
T b f ( )d b ( ) b f ( ) sin d


= =

, (18a)
where
2 2
sin 1 1 esin
f ( ) ln
2(1 e sin ) 4e 1 esin

=


. (18b)
In the last equality of Eq. (18a) we have taken advantage
of the following differential property of the geodesic:
sin d d = . (19)
Expressing d in terms of latitude [e. g. Sjberg 2005a,
Eq. (19)], we finally arrive at the solution
2
1
2 2
2 1
f ( ) sin
T b ( ) b ck d
g(x)

, (20a)
where
( ) ( ) ( ) ( )( )
2 2 2 2 2 2
g x x 1 e 1 x 1 c 1 e c 1 x
l l
=
l l
l l
,

(20b)
2
k 1 e = and x cos = .
Fachbeitrag L. E. Sjberg, New Solutions to the Direct and Indirect Geodetic Problems on the Ellipsoid
4 zfv xxx/2006 131. Jg.
Alternatively, Eq. (20a) can be written
( )
1
2
x
2 2 1
2 1
x
f (x) 1
T b b ck dx
g(x)


=

, (21a)
where
2
1 2 2 2
2 2
1 1 1 e 1 x
f (x) ln
2(1 e e x )
4e 1 x 1 e 1 x

=


. (21b)
In the above equations
2
b a 1 e = ,
1
and
2 1
are
the latitudes at the intersections of the meridians and the
geodesic,
i i
x cos = , ( )
i i
arcsin c/ cos = and
2 2
i i i
cos x / 1 e (1 x ) = . (22)
The integral terms of Eqs. (20a) and (21a), of order e
2
,
are suitable for numerical integration. Note that
1
f (x) is
well conditioned for any x in the interval 0 x 1 with
1
f (1) 1 = .
If
1 2
= , Eqs. (20a, b) and (21a, b) are modified to
max
2 2
1
0
f ( ) sin
T b ( 2 ) 2b ck d
g( )

(23)
and
( )
1
max
x
2 2 1
1
cos
f (x) 1
T b 2 2b ck dx
g(x)

, (24)
where
2 2
max
cos kc/ 1 e c = and
1
c sin = . (25)
5 Numericalexamples
We will present three numerical examples on the direct
and indirect problems and area computation on the ellip-
soid with a = 6378 km and e
2
= 0.00694.
Example 5.1:
a) Let P
1
be defined by
1 1
0 = = , s = 1000 km and
1
60 = . Compute P
2
and
2
.
Solution: From Eq. (2) we get Clairauts constant:
1 1
c cos sin 0.866025 = = .
By iterating Eq. (3b) and (4) we obtain
(0)
2
0.078153 = ,
(2) (3)
2 2 2
0.078524 = = = , which yields
2
4 2956 .81 = ' '' .
Then the longitude of P
2
is given by Eqs. (6a) and (6b),
yielding
2
0.136239 7 4821 .45 = = ' '' .
b) Let P
1
and P
2
be defined from a). Compute s and
1
.
Solution: First c is iterated twice by Eqs. (11) and (12)
to 0.866025. This yields
1 1
arcsin(c/ cos ) arcsin0.866025 59 5959 .83 = = =

' '' .
Then the arc length is determined by Eqs. (3a) and (3b) to
s = 999.9999979 km.
Example 5.2:
a) Let
1 1 2
0 = = = ,
2
50 =

. Compute s and
1
.
Solution: First c is determined by Eqs. (14a) and (14b).
As the integrand of Eq. (14b) becomes singular at
the upper integration limit, it was approximated by
2
1 c 0.005 . After 2 iterations c = 0.999403 had sta-
bilized. Then the arc length and the azimuth at P
1
were
computed by Eqs. (16) and (2) to s = 5557.26246 km and
1
1.536261 88116 .62 = =

' '' , respectively.
b) Let P
1
, s and
1
be given according to a). Compute the
coordinates of P
2
.
Solution: First the latitude of P
2
was searched by Eq. (4),
which did not converge. Then, by assuming that also
2
0 = , Eq. (10) yielded
2
50 .00 00 .00 =

' '' . Then, using
Eq. (16), we obtained s = 5557.26247, which value thus
agrees with its given value in a). Thus we have verified
that P
2
is located on the equator, and the direct problem
is solved.
Example 5.3:
Compute the area under the geodesic limited by the me-
ridians through P
1
and P
2
in Example 5.1.
Solution: The azimuths are given by
1 1
arcsin(c/ cos ) 1.04719674 = = and
2 2
arcsin(c/ cos ) 1.05257549 = = .
From
2
b a 1 e = one obtains the approximate area
2
1 2 1
T b ( ) 217283.20 = = km
2
.
Furthermore,
1
0 = and
( )
2
2 2
arctan tan / 1 e 0.0787968 = = , which
inserted into Eqs. (20a) and (20b) yields
( )
( )
2
1
2
f sin
T bck d 2.36
g

= =

km
2
and, finally,
1 2
T T T 217285.56 = = km
2
.
Fachbeitrag Lars E Sjberg, New Solutions to the Direct and Indirect Geodetic Problems on the Ellipsoid
5 131. Jg. xxx/2006 zfv
6 Concludingremarks
We have demonstrated how to solve the direct and in-
direct problems on the ellipsoid by adding the strict so-
lution for the sphere and an ellipsoidal correction deter-
mined by numerical integration. By employing numerical
integration, the routines of which are usually available in
current computer software like MATLAB, the problems
of classical geodesy are easily solved to desired precision.
Remaining problems to be solved are related with the case
that the two points of interest on the geodesic are located
at nearly the same latitude, implying that the proposed
solutions for the direct and indirect problems become nu-
merically unstable.
Acknowledgement
The author is indebted to Dr. Ming Pan for the numeri-
cal computations.
References
Heck, B.: Rechenverfahren und Auswertemodelle der Landesvermes-
sung. Herbert Wichmann Verlag, Karlsruhe, 1987.
Klotz, J.: Eine analytische Lsung kanonischer Gleisungen der Geod-
tischen Linien zur Transformation ellipsoidischen Flchenkoordina-
ten. Deutsche Geodtische Kommission, Mnchen, Reihe C, Nr. 385,
1991.
Schmidt, H.: Lsung der geodtischen Hauptaufgaben auf dem Rota-
tionsellipsoid mittels numerischer Integration. ZfV 124: 121128,
1999
Sjberg, L. E.: Determination of areas on the plane, sphere and ellipsoid.
Survey Review, 2005a (Accepted.)
Sjberg, L. E.: Precise determination of the Clairaut constant in ellipso-
idal geodesy, 2005b (Submitted to Surv. Rev.)
Authorsaddress
Lars E Sjberg
Royal Institute of Technology, Division of Geodesy
SE-100 44 Stockholm, Sweden
Phone: +46 8 790-7330, Fax: + 46 8 790-7343
sjoberg@geomatics.kth.se