- 5 
of interpretation. It was once stated that for publications of studies 
using the same statistical criteria, a 5% probability level, and the 
same general subjects, one out of twenty would contain erroneous infor- 
mation. The problem is knowing which are the correct reports and which 
are the incorrect reports. 
Aside from the enthusiastic statements regarding remote sensing 
capabilities there are two basic problems with respect to detection 
reliability: a) what does the statistical significance of published 
reports really mean when the critical need is to find and interpret all 
damages, and b) the level of accuracy in interpretation that each report 
indicates was achieved. Should interpretation systems accept results if 
they are 60% correct, or should they be 80% correct, 95 or 100% correct? 
It is recommended that a method be developed to provide detection re- 
liability estimates, and guidelines be stated to indicate the level of 
accuracy that is needed for results to be acceptable to the user com- 
muity. 
D. Photo Interpretation Materials 
Because there is a continuing debate concerning what is the 
best method, and which materials should be used, a study was done to 
test the differences in interpretation results when dead trees were 
interpreted on 1:63,360 color-infrared, and normal-color paper prints 
and duplicate diapositives using a 2-power pocket stereoscope and a 
Zoom 240 magnifying stereoscope. Ground truth was accepted from the 
counts made on 1:4000 scale color-infrared air photos using a 2 x 
stereoscope. Dead tree counts were made on randomly selected plots 
accounting for 10 percent of the total area involved. Results indi- 
cated that the normal color photos at a scale of 1:63,360 could not be 
used for any reliable level of photo-interpretation of randomly dis- 
persed single, or small clumps of dead defoliated trees (Damage Type IB) 
but they could be used equally well for the delineation of the large 
extensive areas of dead, defoliated trees (Damage Type IC). 
Comparison of the results of the color-infrared photos showed 
a marked tendency to over-estimate the number of dead trees on the 
duplicate diapositive when the Zoom 240 stereoscope was used with 15 x 
magnification and to omit observations on the paper prints when the 2 x 
stereoscope was used. Results comparable to the control counts were 
obtained only when the 2 x stereoscope was used on the 1:63,360 dupli- 
cate diapositives. Since it is well known that duplicate diapositives 
tend to emphasize photo contrasts and mask the subtle variations, one 
could expect similar or better results if the original diapositives were 
used. However, the above results suggest a relationship between the 
optical interpretation equipment and the photo product used during the 
interpretation. These relationships should be subject to critical in- 
yestigations. Until the best system is defined, it is suggested that