universal program for adjustment of any geodetic network
TRANSCRIPT
測 地学会誌,第26巻,第3号
(1980),147-171頁Journal of the Geodetic Society of Japan
Vol. 26, No, 3, (1980), pp. 147-171
Universal Program for Adjustment of Any Geodetic Network
(PAG-U)
Takehisa HARADA
Pacific Aero Survey Co., Ltd.**
(Received July 2, 1980)
測地観測量網平均汎用 プログラムの現状
パシフィック航業株式会社 原 田 健 久*
(昭和55年7月2日 受理)
要 旨
測 地観測 量網 平均汎用 プログラムに関する最初の報告を国土地理 院英文報 告(第12巻,1966)に
書 いてか ら既 に14年 の才月が流れた.こ の間計算機の発達 と学の進歩 に応 じて プログラムには何回
も改良が加え られた.大 きな改良については その都度報 告させて いただいた.し か し報告洩れにな
っている事柄 で,し か もこの プログラムに対す る 理解を深め る上では極 めて 重要な事項が幾つかあ
る.例 えば辺長と方向や 角が一緒に観測され ている場合には,な ぜ辺長 の観測方程式の左辺を単な
るv(s)で な くv(s)ρ"/sに す る方が良いのか,と か,未 知点の登録順序についてのサ ツマ芋輪 切 り
論等で ある.い ろい ろな意味で このプ ログラムの改良 も一段落の状態 にある今,1979年 春の学会に
おける 「1点固定のFree Network解 」の報 告と併せて,こ れ までに書 き洩 らして きた諸 々の事 に
ついて述べ る.
ABSTRACT
Fourteen years have elapsed since the first paper on the Universal Program for ad
justing strictly any geodetic network, which consists of four kinds of observations such as side length, azimuth, angle and direction, was presented in the Bulletin of the Geo
graphical Survey Institute, Japan (vol. 12, 1966). The program had been frequently improved according to the development of computer and the advancement of science. Important improvements had been reported in the Journals of the Geodetic Society of Japan every time they were added to the program. There are, however, some important remarks which have not been mentioned in these papers. Nowadays, the improvement of the program seems to be sluggish. Under these circumstances reviewing total flow of the program including these remarks will be useful for persons using the program or considering introduction of the program.
Main features of the program are as follows
1. Available for any reference ellipsoid.
2. Available for the geodetic networks located anywhere on the earth.
3. There is no restriction in number of points.
It depends only on the ability and capacity of computers.
*昭 和54年5月 日本測地学会 第51回 講演会にて発表.**2-13-5,Higashiyama,Meguro,Tokyo,Japan
148 Takehisa HARADA
4. There will be no restriction in size of network, if maximum length of side is less
than 1,000 km. Of course, no matter how small the size is, there will be no
trouble.
5. There will be no restriction in any kind of survey and figural shape of network.
6. Blunders in input data can be checked briefly on the output printed after the
short-time test computation.
7. An advance estimate on the strength of a geodetic network can be obtained by
output-printing diagonal elements in the weight coefficient matrix of all unknown
values (east-west and north-south components in small increment to be added to
approximate position of every unknown point).
8. It is a matter of course that the network can be adjusted strictly by means of
the method of least squares. Network can be naturally adjusted according to
individual preferences, such as multi-fixed points or one-fixed point and one-fixed
azimuth etc..
9. In addition to the conventional methods mentioned above, a new technique of
Free Network solution developed recently is also available. It is based on that
the positions of all points are unknown in the Free Network solution. In this
program both methods are available by which sum of square of each displace-
ment vector for all points becomes minimum and sum of displacement vectors of
all points becomes zero.
10. If flexibility still remains in rotation and size of figure despite the fact that the
position of any single point in the network is known, the method by which sum
of square of each displacement vector becomes minimum can be applied in this
program.
11. Rigorous weight of every observation can be decided automatically according to
its accuracy.
12. Standard deviation of the position of each point found after net-adjustment can
be shown with an error ellipse in addition to both components in east-west and
north-south.
13. Displacement of every point is shown with a vector and both components in east-
west and north-south.
14. Any system of plane rectangular coordinates can be usable, if it is based on the
Gauss-Kruger's conformal projection principle.
15. Final result table for each point will be automatically made including many kinds
of geodetic quantities i, e. coordinates of the points, contraction ratio and the
direction of north at the points, azimuth, direction and distance of every geodetic
line connecting between each of the points and other surrounding points. Of
course some of them are expressed with both values on the ellipsoid and on the
plane. The final result table also contains a few supplemental elements by which
distances and directions can be easily transformed from the ellipsoid to the plane
Universal Program for Adjustment of Any Geodetic Network 149
and vice versa.
16. In case that high precision printing is desired in the output, numerical values of
longitude and latitude can be printed to 6 digits of decimals in arc second and
distances can be printed to 0.1 mm.
17. Radiating sides from a central point can be formed automatically refering code
numbers of the central and other surrounding points in input data. Both old and
new values and variances are shown for these radiating side-lengths and for
horizontal angles formed by the neighboring sides at the center. Especially their
standard deviations can be computed rigorously for side-lengths and horizontal
angles, even if they are not observed directly.
18. The local crustal strains such as rotation, dilatation, major and minor principal
strain axes, and maximum shear strain can be obtained by comparing new tri
angle with old one formed based on code numbers of three vertices in input data.
19. Computation and operation of matrices in this program can be conducted very
fast and efficiently due to the fact that zero-parts in every matrix are completely
neglected in any step such as in storing, adding, multiplying and introducing of
inverse matrix.
I. Outline
Geodetic surveying is concerned with two types of surveying, one for heights and
the other for horizontal elements. Concerning the latter, this program performs precise
adjustments of any geodetic network regardless of its shape or of its location on the earth
by using longitudes and latitudes as expressed by continuous coordinates over the earth's
surface and, in addition, for areas that had been surveyed earlier before the latest sur
veying, it computes land movements and ground strains that took place during the inter-
vening period.
1. Two Cases of Application
While a more detailed flow diagram is given in Figure 13, Figure 1 as shown below
indicates the two cases of application of this program. In order to determine land move
ments, both old and new longitudes and latitudes must be known for each point. There-
fore, when old data are not available, adjustment of a geodetic network is required as a
case shown on the left side of the flow diagram in Figure 1. When both old and new
data are available, land movements can be determined without performing such adjustment,
which is the case shown on the right side of the diagram. When old data are not avail-
able, the geodetic network adjustment alone is performed without computing land move
ments and crustal strains.
150 Takehisa HARADA
Fig. 1. Rough flow of the Program.
2, Outline of Program Structure
The program is structured as an overlay system composed of one main program and
11 sub-routines.
Program Language: FORTRAN IV
3. Computing Speed
Additions and multiplications of matrices take up the bulk of computing time . Figure 2 is a typical array of a matrix with sizeable zero areas in the upper right and the lower
left as represented by A and B respectively in the figure. As the number of points in-
creases, the zero areas rapidly grow in size. And unless something is done, such meaning-
less calculations as (X) +0, (X) X 0, 0+0, 0 X 0, should sharply increase in number adding
wasteful time simply to extend the total computing time.
Therefore, this program totally eliminates both A and B areas , which fall outside of
Fig. 2. An example of matrices in computation.
x : significant values
blank space: zero
Universal Program for Adjustment of Any Geodetic Network 151
the area enclosed by the broken lines, from calculations as well as from storage.
This reduces the total computing time significantly and thus makes it possible to
perform calculations for precise adjustments of a large geodetic network involving enormous
numbers of points, including computations of inverse matrices, which formerly could not
be accomplished due to the tremendous time required. And it also permits evaluation of
accuracies in various terms for the whole geodetic network, which would be otherwise
impossible.
4. Observations
The geodetic network adjustment involves the following four types of observations.
a. Side length
b. Azimuth
c. Direction observation
d. Horizontal angle
These must be expressed in values as projected on the surface of the reference ellip-
soid.
a. Side Length
This is defined as the length of a geodetic line connecting two points on the reference
ellipsoid. It is not a raw measurement value of slope distance but a value corrected for
curvatures of light and electromagnetic waves and heights of observation points.
b. Azimuth
This refers not to astronomical azimuth AQ but to geodetic azimuth Ag. The geodetic
azimuth, however, cannot be directly measured in observation and in application, there-
fore, Ag is obtained from the observed astronomical azimuth as modified by means of the
following Laplace's equation.
Ag=Aa-(ƒÉa-ƒÉg) sinƒÓ-tan h(sin A-cos A) (1)
where and are vertical deviations of the point which are given as follows.
•¬
where (AQ, pa), (Ag, cog) are astronomical and geodetic longitudes and latitudes of the point and
h represents the height angle of the azimuth point as seen from the observation point.
Therefore, the application of azimuths in geodetic network adjustment requires that
the astronomical azimuth and astronomical longitude and latitude of a certain point should
be known by observation and so be the geodetic longitude and latitude of the point.
For such an area where geodetic surveying has never been conducted before, since
geodetic longitudes and latitudes of each point are not known, an adjustment should be
152 Takehisa HARADA
performed initially without azimuths to obtain approximate geodetic longitudes and lati-
tudes of each point and then to find geodetic azimuths by Equation (1). Only after that,
the adjustment is to be performed using the azimuths.
c. Direction observation
One pair observation :
After fixing the horizontal graduated circle of the theodolite at a certain point, several
points in the vicinity are observed consecutively for their respective directions and their
measurements are read out from the horizontal graduated circle. In the next step, as the
telescope is moved back, the points are observed for directions this time in the reverse
order to return to the original position. The two read-out values thus obtained in both
ways, forward and backward, for each direction are averaged to complete one pair obser-
vation.
By changing the fixed position of the horizontal graduated circle, this one pair obser-
vation is repeated for a certain number of times. Observed values are averaged for each
direction and the resulting values are applied as observed values of directions. In PAG-U,
the value for one arbitrarily chosen direction is reduced to 0°0'0" and the rest of the
values are converted accordingly to form a new series of values which are used as input
data.
d. Horizontal angle
This is an angle formed by two observed directions. The value is obtained as the
difference between two directions. When the two directions are observed each with a
weight of p, the weight of the angle is p/2.
5. Geodetic Network Adjustment
Geodetic Network Adjustment:
The values of three interior angles of a plane triangle as measured in observation,
for example, do not usually add up precisely to 180°0'0". This is because of errors in
observation, and raw observation values do not provide any geodetic configuration of
primary significance that they are supposed to represent. In order to obtain a geodetic
pattern of primary significance, it is necessary to modify each observation value in steps
in a manner not to contradict each other. This process is called the geodetic network
adjustment.
Geodetic Network Adjustment by the Least Squares Method:
This is the most precise method for geodetic network adjustment. When v~ is assumed
as a small correction value to be added to each observed value and pb as a weight of
observed values, each value of vti is determined in a manner to satisfy the following formula.
•¬(3)
Universal Program for Adjustment of Any Geodetic Network 153
6. Geodetic Network Adjustment by Observation Equations
In this program, one observation equation is developed for one set of observation and
after theoretically evaluating the weight of observation , each equation is solved by the least squares method so as to satisfy Equation (3).
In so doing, all known points are represented by longitudes and latitudes as already
known but not only that, approximate longitudes and latitudes are assumed for all un-
known points. And observation equations are formed with small correction values as
unknowns to be applied to improve the approximate longitudes and latitudes assumed in
the above.
Observation Equation [1] :
The Observation Equation is defined as follows taking a side length for example.
Assuming the observed value of a side length between Points A and B as Sobs, its small
correction value as vs and the side length computed from the approximate longitudes and
latitudes adopted for Points A and B respectively as Sado (sub ado stands for `adopted')
and small increments to be applied to the adopted longitudes and latitudes of the points
as (ƒÂƒÉa, ƒÂƒÓa) and (ƒÂƒÉb, ƒÂƒÓb), then the observation equation for side length is formed as
follows.
•¬(4)
where coefficients Ca, Cb, da and db are functions of the average latitude of the two points . In practice, Equation (4) above is applied with Sobs transpositioned to make the equa-
tion as follows.
vs=CaƒÂƒÉa+daƒÂƒÓa+cbƒÂƒÉb+dbƒÂƒÓb+ (Sado-Sobs) (5)
I n the actual computation procedure, first, small increments of longitudes and lati
tudes (ƒÂƒÉ, ƒÂƒÓ) and then adjusted longitudes and latitudes (ƒÉ+ƒÂƒÉ,ƒÓ+ƒÂƒÓ) are found after
adjustment of the geodetic network. Subsequently, adjusted side length Sad; is calculated
from them. Finally, vs is obtained from vs=Sad;-Sobs .
7. Free Network Solution
Displacement Vector:
Fig. 3. Displacement of an unknown point.
154 Takehisa HARADA
In Figure 3, Pado is assumed as representing the adopted approximate location and
Pad; as the location after the adjustment of a geodetic network. The arrow from Pado to
Pad) is Displacement Vector as expressed by V. Thus, V represents the actual land
movement from the position as measured in the old surveying to the position as identi-
fied in the new surveying.
The perpendicular components of east-west and north-south of V can be expressed
as NaA cos cv and Mow where
N : Curvature radius in prime vertical
M : Curvature radius in meridian
Free Network Solution:
The conventional method of geodetic network adjustment has been to apply the least
squares method assuming one or more known points in the geodetic network. Recently,
a precise method of geodetic network adjustment has been developed which applies the
least squares method assuming all points as unknown. This method is called the Free
Network Solution.
Geodetic observation values with respect to side lengths, azimuths, directions, hori
zontal angles, etc., do not actually provide any information to fix the location of a geo
detic network. Observations of directions and angles determine its pattern but allows an
infinite number of similar figures. If one or more observed values of side lengths are
additionally available, the dimensions of a figure can be determined but still leaving freedom
of rotation. If azimuths are further given, rotation elements can be accommodated but
Fig. 4. Concept of free network solution is shown in the figure.
Universal Program for Adjustment of Any Geodetic Network 155
parallel movements of the network are left free. Refer to Figure 4. Quadrangle (a) is formed by Pi, P2, Pg and P4 whose locations are already determined by old surveying . Suppose the dimensions of the quadrangle are determined as (b) by the adjustment of the
geodetic network in paucity of azimuth observation in new surveying. In this instance,
if one point and one azimuth are known, the respective points of Quadrangle (b) can be
automatically determined in relation to the reference ellipsoid on which they are based . In the Free Network Solution, however, since all points are unknown , the location of
Quadrangle (b), for example, cannot be fixed on the earth. Now if figure (b) is arbitrarily
overlaid on figure (a) happening to result in figure (c), then a pattern of displacement
vectors for all four points is obtained as shown in figure (c). Depending on the manner
in which figure (b) is overlaid on figure (a), patterns of displacement vectors change in
an infinite variety. In cases like this where all points are unknown, theoretically, point
locations cannot be determined. Despite that, the concept of the Free Network Solution
is to attempt to determine the location of a figure on the basis of one selected pattern
of displacement vectors out of a maltitude of various patterns on certain qualifications.
If the condition assumed for displacement vectors between points is reasonable enough
in physical terms, the displacement vectors chart should be reasonably significant.
This program permits two methods of Free Network Solution assuming,
Square sum of displacement vectors is minimum, V2=min. [2]
and
Total sum of displacement vectors is zero, V=0 [3]
The Free Network Solution in terms of normal equations :
To solve a number of observation equations similar to Equation (5) by the least squares
method, the normal equation as follows derived from Equation (5) must be solved.
BX=L (6)
Fig. 5. Relation between kinds of observations and both methods of free network solutions.
156 Takehisa HARADA
where X represents column vector consisting of small longitude and latitude increments,
B is their coefficient matrix (square and symmetric), and L is real number column vector.
When there are known points in the geodetic network and, therefore, its location is
determined, Equation (6) can be solved as
X=B-1L (7)
In the case of Free Network, however, the determinant of B is
CBI=O (8)
where inverse matrix B-1 has no place and, therefore, Equation (6) cannot be solved.
The statement that a myriad of vector charts can be developed is thus explained in
mathematical terms as above.
Figure 5 shows the types of observations conducted in usually practised surveying.
In the case of a surveying that follows the left side of the diagram, at least either direc-
tions or horizontal angles are measured in observation and so are side lengths and azimuths,
with only freedom left being for parallel movement. Therefore, if the location of one
point in a geodetic network is known, the location of the geodetic network can be deter-
mined. This means that by eliminating arbitrarily selected symmetrical 2 rows and 2
columns in B, the determinant is no longer zero. The defect in rank of matrix B is de-
fined as 2 in this particular instance. As rotation becomes free in paucity of observed
azimuth and sizes of a figure become free due to the lack of observed values of side
lengths. the number of defects in rank of B increases by 1.
The important thing to note is that while the method based on Z Vz=min. applies to
all types of surveying, the method based on E V= 0, on the contrary, is applicable only
when the number of defects in rank is 2. This means, for example, in a figure where
rotation is free, an azimuth is fixed by arbitrarily setting a rotation and after that, the
figure can be moved only in parallel. Then a solution to satisfy E V=0 can always be
found and, therefore, for each varying locking azimuth, there is a solution to satisfy
V=0 and there can be no single solution to satisfy V=0 in a figure where rotation
is free.
This program as shown in Figure 5, makes the Free Network Solution applicable to
all types of conventional surveying. For example, in such a case of triangulation where
neither side lengths nor azimuths are available, all that can be determined from obser-
vations is a pattern of the network which allows as solutions an infinite number of similar
figures subject to free rotations. 4n the other hand, the E V Z = min, method finds only
one solution by appropriately expanding and rotating the figure on the basis of the positions
assumed for each point.
7' Free Network Solution with One Fixed Point
(Supplemented, May 1979)
In the case where one point in the network is known, the Free Network Solution
Universal Program for Adjustment of Any Geodetic Network 157
based on E V2-min. (i. e, the square sum of displacement vectors is minimum) can be
applicable.
Namely, when one point in the network is fixed,
(1) If observed values are available only of side lengths, angles and directions but
not of azimuths:
The geodetic pattern is fixed in terms of size and figure but it permits free
rotation around the fixed point.
(2) If observed values are available of azimuths, directions and angles, but not of
side lengths:
The geodetic pattern allows as solutions an infinite number of similar figures
of all sizes with the same azimuth:
(3) If observed values are available only of angles and directions but not of side lengths nor azimuths:
The geodetic pattern allows free rotation around the fixed point while varying
in size.
Out of these infinite numbers of solutions, the Free Network Solution provides only
one solution so as to minimize the square sum of displacement vectors.
The defects of rank in the normal equations are 1 in (1) and (2) and 2 in (3).
Method of Input Data Preparation:
No special consideration is required but simply to tollow the procedure given in
the manual.
PAG-U recognizes the number of known points and if the number is one and the
side lengths, azimuths, angles, or directions are observed, it performs regular geodetic
network adjustment and, for other cases, the program perceives which of above (1), (2)
or (3) is the case and then automatically performs the Free Network Solution.
8. Final Result Table
Final result tables can be developed for two types of coordinate systems, i. e., Japan's
plane rectangular coordinate system and any arbitrarily chosen coordinate system based
on the same principle (Gauss-Kruger's Conformal Projection Theory) where 2o, con, X0, 10
of the origin and the contraction ratio have to be given.
At each of all points, the vicinity is searched out automatically for such points that
are mutually connected by observations, and the relations between the point and those
connected points in the vicinity are fully represented in a table in terms both of longitude
and latitude coordinates and of the plane rectangular coordinate system as well as con-
versions of such elements as side lengths and directions from ellipsoid to plane and vice versa.
In Figure 6, assuming 0 as the origin of the plane rectangular coordinate system,
A as the point for which a final result table is to be developed, and B as one of the
surrounding points, the meridian can be shown as AN and the geodetic line AB as i.
s represents the distance of AB on the rectangular coordinate system while S is the geo-
158 Takehisa HARADA
detic line length of AB.
In the final result table are represented the following values.
(X, Y )
Contraction ratio: Extent of contraction of a small figure around Point A.
Direction of true north (direction of north reckoned clockwise from the positive direction
of X axis) =Y
(negative in the fig. 6)
Direction
Azimuth=A
Direction= T=A+Y
T-t
t (direction of line segment between A and B on the plane) = T - (T- t)
Distance
Length of geodetic line=S
' S/s
l The length of line segment AB on the plane coordinate system=s=S=-(S/s)
9. Variations of Radial Side Lengths and Sequential Angles around a Point
Refer to Figure 7. Specifying one certain point (P0) and several surrounding points
as P1, P2... P5 clockwise in input data, the following values with respect to radial side
lengths (Sl, S2... S5) and sequential angles (81, 02. .. 85) are represented in output.
With respect to side length :
Old value : Si,old
New value and its standard deviation: Si,new(SD)
Variance : ‡™Si=Si,new- Si,old
Relative variance : ‡™Si/Si
Average value of relative variance : •¬
Universal Program for Adjustment of Any Geodetic Network 159
Fig. 7. Radial side lengths and sequential
angles around a point.
With respect to angle:
Same as for the side length in the above except that S is substituted by d.
The important thing to note here is that these new values of S and 0 are calculated
in relation to a most precise possible location of each point as adjusted even if the side
lengths and angles are not directly measured by observation and that standard deviations
of new values are presented as results of strict computations to indicate levels of reliability
of these values.
10. Location of Point
Error Ellipse:
The accuracy of the location of an unknown point obtained from the network ad-
justment is usually expressed by standard deviations in terms of centimeters in the north-
south and east-west directions. Strictly speaking, however, it should better be defined in
terms of an error ellipse for the reason that the accuracy of the position of a point is
not equi-directional by nature, which means a point location can be determined more
accurately in one direction than in another depending on the figure of a network, method
of observation and its accuracy, and standard deviations due to directions are distributed
Fig. 8. Error ellipse of most reliable position found after net-adjustment for an unknown point.
160 Takehisa HARAVA
in a waisted figure like a cocoon surrounding the point . The maximum value occurs on
one direction of the perpendicular with the minimum on the other direction as represented
by a and b respectively in Figure 8. Since it is rather cumbersome to draw a waisted
figure, it is usually approximated to an ellipse with a and b as axes . Called 'the error ellipse', it is commonly used as a means to show directional variations of reliability of a
point location.
With respect to unknown points, the following values are expressed . (Refer to Figure 8.)Longitude and latitude of approximate position of P ado : ƒÉ, ƒÓ
Small increment for longitude and latitude : ƒÂƒÉ, ƒÂƒÓ
Longitude and latitude of most precise possible position of Pad;
(adjusted) : ƒÉ+ ƒÂƒÉ , ƒÓ+ ƒÂƒÓ
East-west component of standard deviation of Pad; position:
North-south component of standard deviation of P ad; position:Azimuth of major axis of error ellipse: A
Length of major axis of error ellipse :
Length of minor axis of error ellipse : b
East-west component of displacement vector: NƒÂƒÉ cos ƒÓ
North-south component of displacement vector : MƒÂƒÓ
Azimuth of displacement vector: A
Value of displacement vector : V
11. Strain
Strain Ellipse :
The method to express horizontal ground movements in terms of displacement vector
of each point is good so far as it is visual and easy to understand . But as pointed out earlier in Section I-7, charts of displacement vectors come in a variety of forms and fall
short of primary significance. This is a major shortcoming of the method to represent
horizontal tectonic movements in terms of displacement vectors . The strain method is another way to express horizontal changes of the ground by taking note of the changes
in the pattern of a geodetic network . Refer to the left figure in Figure 9. Let's assume a circle with point Po at the center
drawn on the ground at the time of the old surveying . Supposing the ground is subject to compression force in the direction of PoB, the circle distorts in shape as time lapses and land forms change. As a result, the figure turns out to be an ellipse in the new
surveying as shown on the right of Figure 9 . Let's take two axes, P0C and P0D , for example, that crossed each other perpendicularly at Po in the old surveying . In the new surveying, the two axes are no longer perpendicular to each other as seen in the ellipse
shown on the right in Figure 9. Among an infinite number of pairs of such perpendi-cular axes, there is one pair of axes that remain perpendicular . Such pairs of axes are (POA, PoB) and (P '0A', P'0B') as shown in the respective figures, They correspond to the
Universal Program for Adjustment of Any Geodetic Network 161
Fig. 9. When horizontal compression force works uniformly over the ground, a circle (on the left) drawn on the ground changes in form to an ellipse (on the right) after lapse of time.
major and minor principal axes of the ellipse. The rates of elongation or contraction of
the radiuses of both axes of the ellipse from those of the original circle are defined as
strains of major and minor principal axes of a strained ellipse, and they are usually ex-
pressed in the unit of 10-6 (micro-strain). The perpendicular pair of axes (POC, POD)
which are just in the middle between the two principal axes in the original circle loses
perpendicularity to the greatest extent in the strained ellipse. Let •¬C'PO'D' be right
angle -ƒÓ, then 0 as expressed in radian is defined as maximum shear strain. The word
shear implies movement of scissor blades. If a fault movement is anticipated for the future,
faults are likely to occur in parallel with POC or POD. The increase of area in the ellipse
from the original circle is called dilatation. The values of these various strains are not
independent from each other. The difference between the two principal axes is equivalent
to the maximum shear strain and the sum of the principal strains to dilatation. In prac
tice, instead of drawing a circle on the ground, the same changes in figure can be known
from the change in the figure of triangle formed by three points, P1, P2 and P3 which
were located on the periphery of a circle in the old surveying and, therefore, values of
above strains can be determined by computation for each triangle.
This program computes rotation of the region concerned with each nominated triangle,
dilatation, directions and values of strains in the major and minor principal axes and
maximum shear strain. [4]
By a combined use of the strains of primary significance and the displacement vectors
which are easy to see, the ground movements can be estimated more accurately.
12. Q (for evaluation of the strength of a geodetic network)
There should be various methods to evaluate the strength of a geodetic network but
the important thing is whether it provides values most relevant to evaluation of the ac-
162 Takehisa HARADA
curacy of the determined position of each point.
There are three factors that affect the accuracy in determing the position of a point . The first is the pattern of a geodetic network and the second is the type of observation
to be conducted, the third being the accuracy of observation . The last factor, namely, the accuracy of observation depends , on the instruments to be used and, therefore, it is left out of the consideration of the strength of a geodetic network.
In this program, the following values which indicate reliability of the position of each
point in a most straight-forward manner, or, in other words, the values which provide a
standard deviation of an unknown quantity when multiplied by standard deviation of an
observation with unit weight mo are obtained as •¬ (east-west component) , •¬ (north-
south component), and further compounded with east-west and north-south component
in the formulae below, the resulting values are shown in one table for all unknown
points.
•¬
The smaller the values in the above, the better the positions of points can be deter-
mined.
As for such points where Q values are too large, since it means either the array or
pattern of points is not set properly enough or observation is too weak, it should be considered to reselect points or to strengthen observations.
Note : As obvious from (9) above, the true value of an unknown that this program is
concerned with is not OA but OA cos cp.
II. Important Notes
There are two important things to bear in mind in preparation of input data. Four
additional points, though of lesser importance, are also made in the following.
1. Registration Sequence of Unknown Points
when the number of points that form a geodetic network is small, the computing
time is short and it does not matter in what sequence unknown points are registered.
But when the number of points amounts to tens and hundreds, the registration sequence
of unknown points makes a significant difference in the computing time. The good under-
standing of this section should help to plan for the shortest computing time.
As mentioned in Section I-3, the bulk of computing time is concerned with matrices.
And most of the matrices involved are band matrices with numeral values occurring in
a diagonal belt as shown in Figure 2. Computing time depends solely on the width of
Universal Program for Adjustment of Any Geodetic Network 163
Fig. 10. Approximate longitudes and latitudes of unkown
points should be inputed for the purpose of saving
time in computation in order of lemon slices as shown
in the figure. The figure encircled by the dotted line
is an enlargement of the portion of the network having
point No. 50 at center.
this zone. The smaller the width is, the shorter is the time required for computation.
Figure 10 shows a geodetic network. In this particular geodetic network, a number
is given to each point. Please note here that these numbers are serial numbers given as
unknown points are read for their longitudes and latitudes, and they are not related in
any manner to the identification numbers of points read at the beginning of input data.
In all matrices, composing elements are arranged in the sequence of these numbers which
are independent from the identification number of the points.
Unknown points should be registered in a sequence as shown in Figure 10. It looks
like a section of a lemon slice and it is important to cut it into as many such slices as
possible and to arrange them in an orderly manner from left to right and so on (or from
top to down and so on), what is good about this is that in Figure 10, if paying atten-
tion to Point No. 50, for example, 3 lengths, (50-42, 50-43 and 50-51 as shown in solid
line), one angle •¬43, 50, 51 and three directions (shown by arrows pointing to 59, 58 and
42) are observed. Therefore, the unknown points that relate to Point No. 50 occur among
the points from Point No. 42 through Point No. 59. This leads to
59-42=17 (10)
The number acquired in the above formula controls the width of the band in the form
as shown in Figure 2 in relation to Point No. 50 in the band matrix. Such numbers as
obtained in (10) above that control the width of a band matrix can be assumed for every
one of all points that comprise a geodetic network. The smaller the maximum of such
numbers is, the shorter the computing time becomes. If the above procedure is followed
in arranging unknown points, it should accomplish the purpose of reducing the comput-
164 Takehisa HARADA
Fig. 11. In such case as having branches in the network
lemon slices still should be cut without considering the
branches as shown in the figure.
in time.
The aforementioned analogy of a lemon slice to illustrate the arrangement of un-
known points is applicable to any geodetic network regardless of its pattern . Even for
a network with a pattern that branches off, as shown in Figure 11 , it still is comparable to a lemon which is sliced into segments a, b, c, .. , n, 0.. , and, importantly, for where
it branches off, d, d'; e, e'; ..., they are given sequential numbers as if they were con
nected.
2. Weight of Observed Values
Principle:
Observed values should be weighted in inverse proportion to the square of standard
deviations. This is a very basic principle of the least squares method. From this prin
ciple are derived the following.
When the instruments and the method used in observation are identical , the weight
is proportional to the number of repetitions of observation.
The weight of an angle is 1/2 of the weight of the directions that form the angle . (Refer to I-4-d.)
Unit Weight Observation:
It is widely practised that by specifying the weight of one certain observation as 1, the weight of another observation as p is determined according to the above principle . The first thing to be found after least squares computations is the standard deviation of
this unit weight observation.
Universal Program for Adjustment of Any Geodetic Network 165
Case of simultaneous solutions of different kinds of observation values by least squares
method:
Generally speaking, when the least squares method is applied simultaneously to
observed values of different natures (time, mass, length, etc.), each type of value is given
a weight according to the above principle by assuming an appropriate unit weight obser
vation for the respective types. Once observation equations are formed, all it takes is
dimensionless computation and it does not matter in what units the observation equations
are expressed.
When it is considered, however, that the standard deviation of unit weight observation
resulting from computations is something like an average of standard deviations of these
different type observations, it is desirable that the standard deviations of these unit weight
observations are similar in value. It means that vtias in the principle formula of the
least squares method (3), i, e.‡”pivi2=min. is made similar in value as well as in the
extent of their distribution. For this purpose, such a manipulation is required that by
multiplying both sides of an observation equation like (5) by appropriate constant k, vl
as in‡”pivi2 =min. is considered in terms of kvi instead of simple vi.
Case in which a geodetic network is comprised by side lengths and other observed
values:
It is most reasonable to express small correction values to be applied to observed values
of azimuths, directions and angles in the unit of arc seconds. Correction values for azi
muths, directions and angles are represented summarily as v(B). When the small correc
tion value v(s) to be applied to the observed values of side lengths is expressed in the unit
of centimeters, v(O) and v(s) become similar in value. However, when considering a geodetic
network in which there exist triangles of various sizes, v(O) does not change much due to
the sizes of the triangles while v(s) is subject to the following well known formula.
•¬
where a and b are constants to be determined by the type of a distance measurement in
strument used. Therefore, values of v (s) range fairly widely. This makes the standard
deviation of unit weight observation resulting directly from the geodetic network adjust
ment rather difficult for understanding. This problem can be averted almost completely
by using the value of {v(s)/S} which is a dimensionless number for v applicable to pv2
=min, instead of directly v(s) in the observation equation of side lengths . At the same
time, since in this instance the values of {v(s)/S} become very small compared with the
values of v(8), it is necessary to multiply this by large numbers like 106 or 106 to make
it similar to the value of v(e).
Also as can be seen from Figure 12, the uncertainty of side length AB is relative to
the direction or the angle as seen from Point C. In this sense, the correction value of
the side length relates closely to the correction of the directions. Hence an idea is obtained
that by selecting p" (1 radian as expressed in arc seconds which is approximately 2.106)
166 Takehisa HARADA
Fig. 12. The uncertainty of side length
AB is relative to the direction or the
angle as seen from point C.
as a large number multiplier as mentioned in the above, v (s) applicable to ‡”pv2 = min, is
applied as follows.
v(s)/ S•EƒÏ" (12)
When ƒÏ" is considered a dimensionless figure, (12) is a dimensionless value, but ƒÏ"
can also be considered a value in the unit of arc seconds and then (12) is expressed in
arc seconds. When viewed in relation to Figure 12, the latter might be better for under-
standing.
As in this case where side lengths and other values are observed together in a survey-
ing, by applying v (s) p"/S instead of v (s) as the correction value for side lengths applicable
to ‡”pv2=min., both tasks of narrowing the width of v(s) distribution and making v values
similar for a geodetic network can be accomplished simultaneously. Furthermore, in cases
where a direction or an angle is chosen for a unit weight observation , if (12) is applied,
the weight of (12) can be evaluated properly with relation to unit weight observation and,
therefore, the standard deviation of unit weight observation can be obtained from the
geodetic network adjustment in a manner perfectly to suit the purpose. Thus, the advan-
tage is comparable to a case of `getting three birds with one stone'.
In this instance, the observation equation for side lengths is expressed as follows by
multiplying the both sides of (5) by p"/S.
•¬(13)
where
Ca' =CaƒÏ"/S, .. .
When side lengths and other values are known by observation, this program applies
Equation (13). Its weight p is obtained from anticipated standard deviation of unit weight
observation of direction or angle, m", and a and b of (11), basing on the principle men
tioned in II-2, in the following formula.
•¬
Universal Program for Adjustment of Any Geodetic Network 167
When side lengths alone have been observed, on the one hand the procedure mentioned
above can be used, and on the other hand such an adjusting method is also applicable
where Equation (5) as expressed in the unit of length is directly employed as an obser-
vation equation, assuming S ,, as an average side length for unit weight observation. Weight p of observed length S is determined as follows.
•¬
The program allows two alternative`methods, i. e., automatic weighting by computer
performing the above procedure or assigning weights to each observed value in input data according to individual data producers' judgement.
3. Reference Ellipsoid
Two types of ellipsoid, Bessel and International (old), are built in the program. Their
specifications are as follows.
Radius of Equator Inverse number of flattening
m Bessel 6377397. 155 299. 15281285
International (old) 6378388.0 297.0
When an ellipsoid other than the above is used, the name of the ellipsoid, radius of
the equator and inverse number of flattening are to be specified in input data.
4. Approximate Longitudes and Latitudes of Unknown Points
When old data of longitudes and latitudes are available for unknown points, they can
be applied as their approximate longitudes and latitudes. Otherwise, approximate values
are assigined to the unknown points. It does not matter how rough the values are but the
same value should never be given to more than one point. However rough the values of
approximate longitudes and latitudes are, by replacing them with the better locations ob
tained from the geodetic network adjustment and repeating the adjustment two or three
times, the values converge in most cases.
5. Test Computation
Large errors in the input data can be checked by test computations. If the 'test' is
specified in the input data, the program refers the values computed from approximate
longitudes and latitudes of each point to all observed values and without performing a
geodetic network adjustment, it prints them out to terminate. If large discrepancies between observed values and computed values are found in the output, it will be known
168 Takehisa HARADA
whether the errors are in observed values, point numbers or longitudes and latitudes.
6. Number of Digits to be Printed out
In regular print-out:
Longitude, latitude...........The unit of second and 5 digits of decimals
Distance, plane coordinates. .. mm
Angle, direction.............. The unit of second and 3 digits of decimals
When observations are very accurate (where, for example, distances are measured to
0.1 mm), or the average observed distance is 1 km or less:
Longitude, latitude........... The unit of second and 6 digits of decimals
Distance, plane coordinates.. . 0.1 mm
Angle, direction.............. The unit of second and 3 digits of decimals
# Conditions of input data for print-out of high accuracy.
Print-outs of high accuracy can automatically result in the following cases.
(a) When assumed (or old) longitudes and latitudes of unknown points, or longitudes and latitudes of known points include one or more values that have a figure other
than 0 on the sixth digit of decimals.
(b) When the average value of observed side lengths is 1 km or less.
III. Flow Diagram of the Program
Figure 13 is a detailed flow diagram of this program. It shows how the points made
in Chapters I and II are incorporated in the program.
IV. Before Preparation of Input Data
This program has many ramifications, and which branch to take at each decision
point is determined by the instructions specified in the input data. Therefore, all necessary
instructions to choose between alternatives must be given in the input data.
Following are the alternatives to consider.
a. Whether it is test or real computation. (See II-5.)
b. Selection of a reference ellipsoid (See II-3.)
c. Whether to perform a geodetic network adjustment or to compute ground movements
alone without a geodetic network adjustment. (See I-1.)
d. Whether to produce a final result table after geodetic network adjustment or not .
(See I-8.)e. To assign weights to observed values according to individual input data producers'
judgement, or to have a computer do that automatically. (See II-2.)
f. When weights are assigned automatically, which to apply for unit weight observations,
Universal Program for Adjustment of Any Geodetic Network 169
average side length, directions or angles. (See II-2 .)
g. When only side lengths are observed, which to apply for unit weight observations , average side length, directions or angles . (See II-2.)
h. Are there any known points? When there is no known point, the Free Network
Solution must be applied.
i. In the Free Network Solution, is it to be solved based on?:
‡”V2=min., or
‡”V =0, or
Both (‡”V=0 and ‡”V2=min.)
(See I-7.)
j. Whether or not the prior evaluation of the strength of the geodetic network is the main objective. (See I-12.)
k. Whether to obtain variances of radial side lengths and sequential angles around each
point. (See I-9.)
1. Whether to compute strains, (See I-11.)
References
[1] Japanese text book of Geodesy.
[2] E. Mittermayer: A Generalisation of the Least-squares Method for the Adjustment of Free Networks . Bull. Geodesique, 104, 1912, 139-157.
[3] T. Harada: The net-adjustment ‡”V=0 is obtained automatically in the improved Universal Program . Journ.
Geod. Soc. Japan 14, 156-158, 1969.
[4] T. Harada: Calculation of strain was added to the Universal Program. Journ. Geod. Soc . Japan 17, 1-3, 1971.
170 Takehisa HARADA
Fig. 13 (A) Flow of the program.
Universal Program for Adjustment of Any Geodetic Network 171
Fig. 13. (B) Flow of the program.