third, fourth, or more gene- 
   
After photography every second, 
rally, atl photo is supplied with a systematic scheme of 
dots. The scheme may cover the photo totally or partly. In the 
latter case a central part is to be preferred. According to 
terminology of statistics the photos supplied with scheme are 
regarded as clusters while the units defined by the scheme re- 
present the elements. It is easy to see that the design is 
unbiassed if a cluster has only one element and it falls on 
the photo nadir. Later it is assumed, however, that the 
clusters have more elements. It is further assumed, that photo- 
interpretation of elements into area classes is made correctly. 
Sampling on truly vertical photograph -" An illustration 
It is assumed that the inventory area consists of four field 
classes as shown in the Figure 1. The elements of a cluster 
can be expected to fall on the classes according to the dis- 
tribution of the respective cluster (photo) area. The sampling 
can be regarded unbiassed if the distribution of photo areas 
into classes is equal to the true field distribution. 
The number of the photos in the illustration equals that of 
the classes. The proportions of the area classes are estimated 
on the phote scale by lengths of a,.b, c, and d as: 
4 4 y 4 
Pa z2.a,/s, Bg: Z b/s, Pc =2Zc./s, and By.” 2. d,/s, where 
i ea subscript for a photo and 
4 
S =Z(a,+b,+c,+d.) 
i i "1 i 
The class lengths on the photo scale are presented in measures 
of datum level (see Fig.) in the following table: 
  
Terrain class 
  
Photo a | b € d 
Length