not used. The functions were derived from National Forest Survey data
collected over large areas. More than 3000 trees were used to derive the RE
functions for pine and spruce. To get the volume of a tree there exist =
in Sweden general functions which use dbh and tree height as independent A
variables.
Ground truth data for the stands were collected in the field by means
of stem counts or strip cruising methods. /2
The following accuracy results for stands were cbtainded:
(Image scale 1:8 000)
X S RMS /3
Mean height weighted by basal area -0.8 m 1.1 m 1.3 m
Mean dbh weighted by basal area -1.5 cm 2.6 cm 3.0 cm
Basal area per hectare -7% 15% 16% /4
Volume per hectar -2% 18% 18%
where X = systematic error (bias)
s = standard error (excluding bias) /5
RMS = Root mean square error (including bias)
The number of trees visible in the photographs (dbh greater than 8 cm)
accounted for 70 % of the field observations. Pine was overestimated
and spruce underestimated.
Finally, an analysis was made of some known sources of systematic errors.
This involved the construction of a statistical model for estimating
hidden trees and their influence on collected data. Some improvements
can be obtained in correcting for the hidden trees. Also a study of the
influende in area determination of the circular plots was made. This
was not covered in the separate accuracy study. As the sample plots
are definied in the image they will vary in size in the terrain. This
will be accounted for in the future.
The largest error source lies in the applied functional relations between
tree heights and diameters at brest height. Correction terms should be
used which account for local variations if general functions are used.
Otherwise local functions have to be derived.
It is also suggested that normal-angle photography would be preferable
to the wide-angle photography employed in this study.
The conclusion to be drawn is that scope exists for increasing the
accuracy of the method studied.
Final remark
This paper is a summary of a report in Swedish /5/.