- 1784 -
This way of evaluation eliminates some of the internal variance
within stands and reduces the effect of the stand boundaries.
The results still remain modest.
An idea of the uses of the results is illustrated by Fig. 3,
where the thematic vegetation map is introduced as an additional
file (overlay) to topographic map, both being in the same coordi-
nate system.
As an intermediate conclusion one could say that the procedure -
applied seems to differentiate between timber types on moraine
soils but requires major refinements for other soil types. Problem
areas to be removed in the continuation involve 1) the seasonal
effect, 2) the lack of thermal information on soils, 3) the in-
sufficient information on the condition of vegetation,and 4) the
lacking knowledge of spectral properties of the classes studied.
4.2 Timber inventory
The timber volume functions derived for Landsat data were tested
in two ways: first by visual checking and, second, by calculating
the correlation between known sample plot volumes and their esti-
mated volumes. On the basis of visual checking, it was obvious
that the best functions separate well the timberlands from open
areas. The more detailed evaluation was done by correlation anal-
ysis which resulted coefficients from 0.80 to 0.90. The best
function, mentioned above, gave a correlation coefficient 0.92.
The results were applied in a timber inventory using a multistage
sampling system. The volume estimates were derived for primary
sampling units by using the volume function. In the first stage
sampling procedure, the variable probabilities (pps) were derived
from those volume estimates. The whole sampling scheme is being
tested in 1978.
The supervised classification of Landsat data was also used for
the extragtion of cutting opportunities in the forestry test site.
The timber land area was divided into five cutting opportunity
classes, and the results of classification are given in Table 5.
TABLE 5. The pixel by pixel performance of the cutting op-
portunity (stocking) classification (Kuhmo test site).
bou Ground |
MS truth |
Landsat | 1 2 3 4 5 Total
~
|
classi- ~~
fication il %
1. Old, high stock-
S.
Water
=
ing timber stands | 75 25 100
2. Old, low stock-
ing and thinning 5 90 3 2 100
stands
3. Seedling stands
and open peatlands | 2 91 7 100
4. Clear cut areas | 10 90 100
The
In
be
in
of
(d)
In
ing
in
mat
var