GVII-68 
PHOTOGRAMMETRIC ENGINEERING 
The results of such comparisons are shown in Table 1. The variance of the 
differences for photo estimates to ground estimates for volume are listed in the 
order of the increasing magnitude of photo scale. This wide range of scales 
appears to indicate no relation of photo scale and of the reliability of volume 
estimates obtained from them. 
Table 1 
Variance of Differences between Photo Volume and Ground Volume 
for Various Photo Scales 
Study 
Photo Scale 
(Representative fraction) 
Variance of differences 
1 
1,200 
86,936 
2 
5,000 
29,000 
3 
7,200 
221,700 
4 
7,900 
56,500 
5 
12,000 
171,400 
6 
20,000 
271,200 
7 
20,000 
3,800 
8 
20,000 
46,000 
Comparison of Tests 
A glance at the variances in Table 1 furnishes no evidence that the use of 
larger photo scales will result in better photo interpretation. If one stops here, 
that would be the logical conclusion. In fact, study 7 shows the smallest variance 
and the smallest scale. One should look further and see what factors, other than 
photo scale, influence the variance of the differences of photo estimate and 
ground estimate of volume. Such other factors are plot size, variability of vol 
ume, quality of aerial photography, paired or unpaired observations, photo 
interpretation techniques and the skill of the photo interpreter. 
Only where all these factors are about the same, can studies be compared. 
There are no acceptable or reliable schemes to make data comparable when 
plot size, variability of volume, photo quality, observations, photo interpreta 
tion techniques and skill of interpreters are not the same. Such comparisons 
selected by the author after careful study, do show smaller variances for the 
larger scales. This is very evident for comparable studies 1 and 3 by Losee (4) 
for scales of 1,200 and 7,200 respectively. Here the variance of the larger 1,200 
scale is 0.39 of the variance of the smaller 7,200 scale. In another case, study 
5 by Dilworth (2) and study 6 by Pope (7) are comparable. Here the variance 
of the larger 12,000 scale is 0.63 of the smaller 20,000 scale. These studies 
indicate that larger scales in aerial photos will improve accuracy of photo inter 
pretation for volume. However, there are no comparable studies between 7,200 
and 12,000 photo scales. But if it is assumed that the reduction in variance is 
at the same rate in this zone, then the data will show the variance ratio to have 
a trend as appears in Figure 1. This assumed trend (dashed line) will give a 
variance ratio for the 7,200 scale of 0.41 and 0.13 for the 1,200 scale. By using 
the 0.39 relation, established by Losee’s studies (1 and 3), the variance ratio for 
the 1,200 scale is computed as 0.16, which is 0.03 higher than the theoretical 
ratio computed from the trend of studies 5 and 6. This represents the trend 
for a given combination of factors that affect variance of differences. 
Studies 2 by Rogers (<?), 4 by Dahl (7) and 8 by Moessner (6) cannot be 
compared with any other studies. The techniques used by Rogers and Moessner 
prevent valid comparisons with other studies. Rogers used Sonne continuous