al 
If 
rest 
ions 
ed 
ry 
east 
The 
lour- 
puter 
lues 
t- 
for 
ected. 
ation 
re l). 
forest 
. which 
f the 
pecies 
onal 
mpos- 
cal 
large 
ses 
-.1761 - 
Type 1 Poplar and Birch 70 - 100% 
Type 2 Poplar and Birch 50 - 70% 
Type 3 Balsam Fir 50 - 90% 
The lesser species included within Types 1 and 2 are white spruce, 
balsam fir and jack pine, and the lesser species included within Type 3 
are white birch and poplar. 
The areas covered by the three broad classes described above are 
presented in Figure 2. Table 2 lists the detailed forest stands inter- 
preted on the airphotos which compose each of these three broad classes. 
It will be noted in the list of components of Type 3 that some white 
spruce stands were included within the balsam fir stands. The reasons 
for this inclusion were the small area of the white spruce stands and 
the difficulties of separating these two species both on airphotos and 
by their reflectance characteristics on Landsat. 
An enlargement of a Landsat image produced directly from magnetic 
tape, illustrated in Figure 3, shows the pronounced differences in 
reflectance values among the forest types. Figure 4 illustrates the 
results of the supervised classification of the Landsat tapes with 
variations in grey shades indicating the colour-coded patterns. As 
can be seen within the analysed image, there are numerous small areas 
that are not classified. This results from the fact that the range 
of reflectance values used to define each of the three forest stand 
types was narrower than the range of reflectance values occurring 
within the type. To reduce the unclassified areas, pixels of similar 
but slightly different stands would be identified and the digital 
values of these pixels would be coded with the same colour as that 
given to the group in which they occur. Stand conditions which are 
completely at variance with those classified would, of course, still 
remain unclassified. 
When the original detailed aerial photographic typing is directly 
compared with the digital analysis results it is evident that the 
majority of the unclassified areas do indeed correspond to species 
not included in any of the three major types. This comparison gives 
evidence that computer analysis may exclude forest cover different 
from the types of interest to a degree similar to that of aerial 
photographic interpretation. 
A comparison between the computer analysis and the generalized 
aerial photographic interpretation reveals, first of all, similarity 
in the total area estimates for the same type in the two analyses 
(Table 3). In general, the location and distribution patterns of the 
types were the same by both methods but the boundaries delineated did 
not match. This lack of fit is to be expected since the level of 
generalization of mapping is not the same for both methods. When