Had the PU areas been determined inconsistantly, the result would have
shown a higher standard deviation. On the other hand consistantly too
high or too low areas would have no influence on the results. When the
values of the standard deviations becomes too high, the method may no
longer guarantee an efficient application. This may be the case when
very small scale aerial photographs or satellite images on which the
interpretation of detail is inconsistant are used. If this is found,
an improvement in resolution of the existing satellite recording system
is recommended. On the other hand, the orthophoto mosaic, scale 1:10,000,
from infra-red photography (1:30,000) was well suited for a consistent PU
area determination.
Assuming that a coefficient of the standard deviation (CVs) - column 7
Table 2 - should not be larger than 13 per cent, according to Hansen,
Hurwitz, Madow (1953), in order to have a reliable estimate of the standard
deviation, the combination of SU's of the 225 ha grid still gives acceptable
results. For 450 ha and 675 ha grid cells the 2 and 5 SU's become doubtful.
As the CVs are calculated from 35 samples, their values are only an
estimate and so the values can be considered as a boundary case.
When a choice must be made from the given combinations, there is a tendency
to select the one with the lowest standard deviftion, as for a certain
standard error, a lower standard deviation results in a lower number of
PU's in the sample. Thé lowest most reliable value is obtained from the
675 ha grid with 35 SU's/PU (8% = 10.8). For a standard error per cent of
5: S% = 5, this would result in
10 8°
2 = 5 PU's and therefore 5 x 50 = 250 SU's.
5
Considering now the 225 ha grid cells with 2 and 10 SU's/PU for 8% = 5,
respectively
2 2
2
ER = 25 PU's (and 25 x 2 = 50 SU's) and ew = 9 PU's
5 5
(and 9 x 10 = 90 SU's), will be involved. It is therefore obvious that
the 225 ha grid with 2 SU's is selected.
As sampling is done with replacements, the same SU data of the reselected
PU’s can again enter into the calculation. For larger areas, reselection of
PU's should not be expected to occur frequently. For the given 10,000 ha
forest of Austria, replacements of selected PU's resulted in data which
are given in Table 3.
Tabl:
Redu
invo
and
grid
225
225
x)
From
smal
with
othe
An €
A-pr
ster
esti
volu
Appa
and
devi
81-1
41-1
The
in t
age-
If a
very
two-
lead
It h
meth
conc
For
limi
list
calc
Tabl