- 38 -
d - To establish the necessary geometric fidelity for a system of world-wide tracking sta -
tions for observing orbiting satellites. The data collected from these satellite observations will
be ideally suited for analyzing gravimetric and related geophysical parameters. These parame
ters, in turn, can then be used for determining the position of the center of mass and the ove -
rail shape of the gravitational field of the earth associated with its mass distribution.
Such results will support the establishment of a uniform system for both geometric and
gravimetric data, which is desirable from the standpoint of basic geodetic theoretical conside
rations and, in practice, necessary to provide information for anticipated navigational problems
in connection with space explorations.
In gravimetric geodesy difficulties in the classical theory are caused by the need for
certàin information which, in practice, cannot be ascertained. The theory postulates that cer
tain groups of observations are executed at the geoid, while in practice they obviously can only
be made at points on the physical surface of the earth. Subsequent reductions depend on know
ledge of the distribution of mass inside the crust of the earth. Further difficulties arise from
the fact that the theory assumes that all masses of the earth are inside the geoid. The process
of regularizing the earth affects the corresponding results in accordance with the postulated ty
pe of mass transport.
New ideas in gravimetric geodesy suggest that the determination of the geoid is an un-
solvable problem and in practice an unnecessary operation. Nevertheless, the dilemma remains
that the paucity and non-uniformity of the distribution of gravity measurements produces errors
larger than those caused by the defects of the theoretical developments. In addition, the corre
lation between the classic geometric triangulation results and the gravimetric data still depends
on certain a priori accepted hypotheses.
The ability to calculate the orbits of artificial satellites in terms of the potential is the
basic significance of such vehicles for world-wide gravimetric geodesy. Because the partial
derivatives of satellite orbit observations with respect to the positions of the specific tracking
stations allow us to determine these positions relative to the center of mass, we are provided
with a link between the gravimetrical and geometrical problems in geodesy. In practice, howe
ver, there are errors in the relative positions of such tracking stations within the individual da
tum. Also, rather strong correlation exists between the large number of perturbations, caused
by the variations of the gravitational field and other geophysical phenomena, and the station po
sition errors. To overcome the difficulties in deriving a solution to the gravimetric problem, it
is necessary to obtain, as a first step, a geometric satellite solution such as was mentioned
earlier.
Considerable interest has been shown in the launching of a series of gravimetric sa -
tellites. Also, valuable gravimetric data has already been obtained by the analysis of satellite
orbits which were observed mainly by photogrammetric cameras [3]. Any forthcoming geode -
tic satellite program will, therefore, include gravimetric satellites with optical beacons for
photogrammetric triangulation in support of orbit determinations. (Compare flow chart, " Geo
metry of Satellite Orbits " .
The processes of data adjustment and analysis in satellite photogrammetry will depend
on the acceptance of a reformation in a third ar&a of activity, which is concerned with modern
mathematical statistics and amounts to a generalization of the classic least squares method. It
is no longer sufficient to consider the purpose of an adjustment as a means of reducing obser -
vational errors, but the ultimate goal is to make a set of measured quantities compatible with
an economized mathematical model. The basic idea underlying this approach is that all quanti
ties used in the construction of a specific mathematical model are assumed to be measured
quantities, which in turn are considered to be samples of stochastic variables. By assigning to
these variables in advance relative variances and co-variances between the limits of zero and
infinity, the least squares method will provide unbiased estimates for all variables, even if