sis
been
or
al
es 1
r
s can
diff-
e
uld
ntem-
tes of
itions
lass-
d
imated
1769 ~
value for these pixels is an average of the reflectance values of the
two types rather than the reflectance value of the type which is
dominant within the pixel.
CONCLUSION
The results from the initial testing of the Landsat forest typing
methodology are promising and warrant the continuation of the research.
The next stage of the work,which will begin in the fall of 1978, is to
test the accuracy that can be achieved when the digital typing of
forest stands is extrapolated to an area of approximately 2,000 km?.
As the Ontario Centre for Remote Sensing acquired its own digital
analysis capability in the spring of 1978, intensive research on this
methodology is now scheduled.
The authors anticipate that the digital analysis of data of Landsat
1, 2 and 3 will prove adequate for a rapid forest typing technique to a
level of detail acceptable to forest managers. It is further anticipated
that the superior data of Landsat D, with its 40-meter resolution and
256 radiometric reflectance levels, will increase the scope of digital
analysis for forest inventory applications. The increased complexity
of data recording will make analysis more critical; indeed, therefore,
the selection of training sites and isolation of the pixels correspond-
ing to those areas will require great care. In addition, increased
numbers of training sites will be required in order to avoid the
exclusive mapping of an unnecessarily narrow range of forest conditions.
December 2, 1977