INTRODUCTION 
This paper discusses a technique of mapping forest stands using 
Landsat data by extrapolating multispectral reflectance values for 
specific types of forest stand which have previously been determined 
by field study of sample locations. Although the research described 
has been performed in the boreal forest of the Province of Ontario, 
Canada, the methodology would appear to be relevant to any boreal 
forest region in the world. 
Forest inventory is conducted on a twenty-year cycle in Ontario, 
as described by Dixon et al, 1965. The primary data source is the 
1:15,840-scale panchromatic aerial photography of the eight regions 
of the province, obtained successively, so that each region acquires 
new coverage every 10 years. The typing is done to a minimum stand 
size of eight hectares. 
At this time there is no thought of replacing the present system 
of forest inventory with the approach discussed in this paper. For 
basic forest inventory data the present system is an effective and 
efficient procedure when compared with extrapolation from ground data 
alone or with the state-of-the-art at the present time of operational 
forest inventory using satellite data. 
However, major changes can, naturally, occur within the twenty- 
year period which render the inventory data outdated for the purpose 
of producing timber volume estimates on,which to base management or 
cutting operations. It can also happen that volume estimates are 
required for areas where no inventory data is available. In such cases, 
a quick.and inexpensive method of obtaining an up-to-date estimate of 
the area covered by the growth of a particular species in a particular 
area is needed. 
As one means of securing up-to-date information, research on the 
use of small-format aerial photography has been in progress for several 
years (Zsilinszky, 1968; Zsilinszky, 1972, Meyer, 1974). An alternative 
method is the digital analysis of Landsat data which is described in this 
paper. 
A method of mapping coniferous and deciduous forests on Landsat 
has been reported by Kalensky et al, 1974. The investigation discussed 
here attempts an extension beyond the separation of broad coniferous and 
deciduous types, to identify forest stands. 
Forest typing using Landsat data is relatively objective. It is 
based on the recording of reflectance values from a perspective suffic- 
iently remote to eliminate the effects of shadow, radial displacement, 
and close-up detail, which influence the recording of reflectance values 
from forest types on aerial photography and are, in fact, essential to 
the discrimination of forest types by the human eye. In addition, 
forest inventory photography is panchromatic, while Landsat data is 
mult: 
of t] 
inte 
inte 
of L 
sele 
to b 
inve 
of t 
asso 
the 
bore 
majo 
are 
(Pic 
popl 
spec 
(Abi 
and 
vali 
of t 
and 
are 
in 4 
not 
exar 
ano! 
inf: 
tha: 
jac! 
WOO 
the 
spe 
The 
Spr 
ref 
is