INTRODUCTION
The most recent classification of Canada's forests was published by
Rowe in 1959 and revised by the same author in 1972. This classification was
based on the work done by Halliday in 1937. Of necessity, these
classifications were based on partly subjective data obtained from a variety
of sources. Such an inconsistent data base may pose serious problems.
In 1971, it was concluded that the uniformity of data provided by
ultra-small-scale (1:160,000) aerial photography could advantageously be used
to revise and update this type of forest classification (Nielsen and Wightman,
1971). A potentially better source of up-to-date information is provided by
sensors aboard satellites such as the Earth Resources and Technology Satellite
(ERTS-1). The present study utilizes these data and represents a logical
extension of the earlier work.
The objectives of the study were to determine whether or not it is
possible to classify forest cover on a regional basis using only ERTS-1
imagery. Because boundaries previously recognized on ultra-small-scale aerial
photography coincided closely with Rowe's boundaries, it was assumed that
Rowe's map could again be used as a basis for comparison. The study was
confined to the Province of Ontario with a total area of about 940,000 knr and
to imagery and interpretation methods normally available to the average user.
METHOD
All imagery of the province with cloud cover up to 10 per cent and
rated "excellent" and "good" (with few exceptions), available since launch to
May 1974, was obtained. These images were classified into "summer" (May to
September) and "winter" (October to April) images. A third group comprised
all colour composites available (mostly combinations of MSS bands 5, 6 and 7
obtained during the winter).
From each one of these groups of prints, a rudimentary mosaic at a
scale of approximately 1:1,000,000 (Fig. 1, 2 and 3) was constructed. World
Aeronautical Chart series maps were used as a working base. The
black-and-white mosaics were constructed primarily from band 5 images.
Although all imagery available was used, many gaps still occurred.
Two interpreters, working independently of each other, and
inexperienced in interpretation of satellite imagery were asked to delineate
what they considered forest-type boundaries. The criterion for^ this
delineation was simple. Without considering areas smaller than 5,000 km, and
without using any interpretation aids, any differences based on tone, texture
or other variables which could be caused by differences in forest composition
were delineated. This criterion was to be applied by the interpreters
independently on all mosaics.
These delineations were then transferred onto a 1:2,000,000 scale
map showing Rowe's forest classification boundaries which are primarily