eral 
| on 
ight 
this 
hen 
it in 
rror 
ived 
own 
ting 
ex- 
hout 
sives 
etric 
stru- 
ct a 
le C- 
. ap- 
y of 
)pe. 
ES 
| que 
ir les 
rte à 
tance 
ique. 
nclu- 
nter- 
e has 
 per- 
shing 
  
  
  
(493) 
to have graphic evidence presented to show that a contour interval which 1s 
3.33 times the mean square error in height is the smallest interval which will 
give a rendering of the topography uninfluenced by the differential uncertainty 
of drawing. In other words, the American practice of planning aerial photo- 
graphy based upon empirically determined C-factors will guarantee the pro- 
duction of contours which can be properly modified by a topographically quali- 
fied operator. To exceed the C-factors, proved by experience, is to cause the 
stereoplotting apparatus to spew forth contours like so much spaghetti which 
cannot be made to resemble the topographic features. 
The statement that the minimum contour interval is 3.33 times the mean 
square error in height leads to several questions. First, can error theory be 
applied strictly to the problem of contour accuracy? Second, what height error 
is significant in the appraisal of instrument performance, — that of stereoscopi- 
cally read spot heights or that of points whose elevations are interpolated from 
stereoscopically drawn contours? 
To the first question one can reply that error theory cannot be applied 
rigorously to the analysis of spot height errors in the stereoscopic model. Gaus- 
sian theory has been introduced in an a posteriori sense by the erroneous as- 
sumption that errors exhibit a normal distribution in the stereoscopic model. 
This has proved to be a false supposition. For example, experience has shown 
that spot height errors on definite points of known elevation, when read in a 
properly functioning stereoplotting instrument, will not exceed a maximum of 
1/. times the contour interval. From this fact the Gauss theory would fix the 
mean square error at ‘/15 times the contour interval. 
Obviously, there is a great disparity between this ratio and the ratio found 
from experience to be associated with the mean square error of points whose ele- 
vations were interpolated from the contours. The contour inaccuracies which 
exist in the final topographic map are of vital interest to the photogrammetrist. 
They may be likened to the body of the beast. The posterior addition of error 
theory analysis resembles the tail and we must not allow the tail to wag the beast. 
The second question is partially answered when we note the difference 
between the mean square error of spot heights in the stereoplotter and that of 
spot heights whose elevations are interpolated from stereoscopically drawn con- 
tours. Field tests of many topographic maps compiled photogrammetrically by 
the U.S. Geological Survey have generally confirmed the C-factors assigned to 
the various instruments. In spite of this satisfying confirmation, test procedures 
might well be improved. Since contours are not drawn by the interpolative 
methods of the plane table topographer, they should not be field checked by 
interpolation. Rather, a net work of spot heights on identifiable points could be 
established by the stereoplotter. Thesc would be checked in the field to deter- 
mine instrument performance. After this check, the instrument itself could be 
used to test contour accuracy, not by redrawing, but by a spot-heighting tech- 
nique which would eliminate continuous notion of the floating mark: that is, the 
operator would set the height counter to the contour value and “bump” the 
mark against the model to check the contour position accuracy. 
Such a test would be comparable to that made by Prof. Finsterwalder when 
he compared multiplex with stereoplanigraph contours. But, it avoids a com- 
parison of instruments and bases the appraisal of plotter-operator combination 
upon absolute ground elevations.