830 
METHODS 
Air Photo Identification Keys 
One part of the argument supporting the probabilistic key approach 
to air photo interpretation is made apparent by focusing attention on the 
problems of constructing a key only from raw photo-descriptive data of 
agricultural crops. As stated in the introduction section of this document, 
identification keys vary in form and style. The most useful keys are 
designed to assure that raw photo data are transformed into a key with a 
minimum of information loss. 
The nature of the raw descriptive data determines to a great extent 
the style of key which must be constructed in order to minimize the loss 
of information. In many instances particular to agricultural interpretation, 
for example, the construction of a probabilistic key is the only approach 
which does not require a simplification (sacrifice of information) of the 
raw photo-descriptive data. In order to illustrate this important point, 
and to promote an understanding of probabilistic keys in general, a hypo 
thetical "case study" example of two types of key construction is presented 
in the following sections. 
Hypothetical Example : The results of a densitométrie analysis of 
known agricultural land use classes is presented in Figure 1. For this 
example, we assume that the results shown in Figure 1 were obtained by 
measuring the positive transmission photographic densities of known fields 
in a training or ground truth area. The data constitute a sample of the 
"tones" which correspond to several land-cover types. For simplicity, the 
five land-cover categories are referred to as "crop" types. 
The plotting of basic descriptive data presented in Figure 1 is a 
prerequisite step common to all key construction endeavors, although 
subsequent events differ according to the type of key being constructed. 
Conventional Selective Elimination Key : If the decision is made to 
construct a conventional dichotomous-type key to the "crops" based on 
Figure 1 data, then the information must first be recognized into a 
comparison matrix (Figure 2). (The comparison matrix form of the data 
contains all of the information found in Figure 1 except for the frequency 
distribution aspect.) Using the comparison matrix, it is a straight forward 
matter to construct a conventional type, selective elimination key. 
Figure 3 illustrates one possible version. 
Even though all the information contained in Figure 2 was used to 
construct the key shown in Figure 3, the resultant "key" does not always 
lead to an unambiguous answer because of overlapping crop descriptions. 
According to Figure 3, a density measurement of .38 on 1000 degree day 
imagery (late spring) and .20 on 2000 degree day imagery (late summer) 
leads to either spring wheat or_ oats. (The designation "degree day" is a 
more precise way to "date" each image according to the current growing 
OCCURENCES